ICSB2008

ICSB2008 The 9th International Conference on Systems Biology The 9 th International Conference on Systems Biology

Abstract book Abstract Abstract book

www.icsb-2008.org www.informtrycket.se Table of Contents

ICSB 2008

Abstract no Page

Plenary Sessions 5 Opening session P-01 – P-02 6 Plenary session 1: Network biology P-03 – P-06 6 Plenary session 2: Understanding and curing diseases P-07 – P-10 7 Plenary session 3: Cell regulation P-11 – P-14 8 Plenary session 4: From cell to organ to organism P-15 – P-18 9 Plenary session 5: New approaches to biotechnology P-19 – P-22 9 Plenary session 6: Genetic variation and evolution P-23 – P-26 10 Plenary session 7: Future challenges P-27 – P-29 11

Dedicated Orals 13 Dedicated session 1-1: Cell-regulation – metabolism DS1-1-01 – DS1-1-08 14 Dedicated session 1-2: Standards and repositories DS1-2-01 – DS1-2-06 16 Dedicated session 1-3: Drug discovery DS1-3-01 – DS1-3-08 18 Dedicated session 1-4: Plant systems DS1-4-01 – DS1-4-06 20 Dedicated session 2-1: Cell-regulation – signalling DS2-1-01 – DS2-1-08 22 Dedicated session 2-2: Modelling approaches DS2-2-01 - DS2-2-08 25 Dedicated session 2-3: Diagnostic markers and complex diseases DS2-3-01 - DS2-3-08 27 Dedicated session 2-4: Microbial systems DS2-4-01 - DS2-4-08 29 Dedicated session 3-1: Cell-to-cell variation DS3-1-01 - DS3-1-09 32 Dedicated session 3-2: Synthetic biology DS3-2-01 - DS3-2-08 34 Dedicated session 3-3: Software tools DS3-3-01 - DS3-3-08 36 Dedicated session 3-4: Model driven experimental planning DS3-4-01 - DS3-4-08 39

Arenas 43 Arena A-01 - A-55 44

Dedicated Posters 61 Dedicated session 1-1: Cell-regulation – metabolism DS1-1-09 - DS1-1-70 62 Dedicated session 1-2: Standards and repositories DS1-2-08 - DS1-2-19 79 Dedicated session 1-3: Drug discovery DS1-3-12 - DS1-3-27 83 Dedicated session 1-4: Plant systems DS1-4-09 - DS1-4-33 86 Dedicated session 2-1: Cell-regulation – signalling DS2-1-09 - DS2-1-126 92 Dedicated session 2-2: Modelling approaches DS2-2-09 - DS2-2-153 125 Dedicated session 2-3: Diagnostic markers and complex diseases DS2-3-09 - DS2-3-24 163 Dedicated session 2-4: Microbial systems DS2-4-09 - DS2-4-45 167 Dedicated session 3-1: Cell-to-cell variation DS3-1-10 - DS3-1-19 176 Dedicated session 3-2: Synthetic biology DS3-2-10 - DS3-2-19 178 Dedicated session 3-3: Software tools DS3-3-09 - DS3-3-62 180 Dedicated session 3-4: Model driven experimental planning DS3-4-09 - DS3-4-25 195 Other session OS-02 - OS-50 200

Author Index 213

ICSB 2008 3

Plenary Sessions Plenary Sessions Opening session Plenary session 1: Network biology

P-01 P-03

Networks from experiments: combinatorial perturbation of Understanding regulatory circuitry through expression- cancer pathways profile phenotypes Sander, Chris Holstege, Frank Computational and Systems Biology, Memorial Sloan Kettering University Medical Center Utrecht, Physiological Chemistry, Cancer Center, New York, United States Utrecht, Netherlands Sessions Plenary We present a novel method for deriving network models from The availability of whole genome sequences and the parallel molecular profiles of perturbed cellular systems. The network development of various high-throughput techniques is making models aim to predict quantitative outcomes of combinatorial it possible to analyze and understand regulatory processes perturbations, such as drug pair treatments or multiple genetic in a systematic manner. In the long-term, this will lead to the alterations. Mathematically, we represent the system by a set development of genome control maps, exhaustive wiring of nodes, representing molecular concentrations or cellular diagrams which describe in intricate detail the role of any processes, a perturbation vector and an interaction matrix. After regulatory factor towards regulation of every single gene. We are perturbation, the system evolves in time according to differential systematically generating DNA microarray mRNA expression- equations with built-in non-linearity, similar to Hopfield networks, profiles of targeted mutations in components of the signaling capable of representing epistasis and saturation effects. For a and transcription machinery in the yeast S. cerevisiae in order particular set of experiments, we derive the interaction matrix to uncover new regulatory mechanisms. A previous pilot study by minimizing a composite error function, aiming at accuracy of which focused on the coregulatory Mediator complex has prediction and simplicity of network structure. To evaluate the shown the feasibility of interpreting such expression-profiles for predictive potential of the method we performed twenty-one structure-function analyses, discovering new regulatory pathways, drug pair treatment experiments in a human breast cancer cell uncovering epistasis and pin-pointing the precise effects of line (MCF7) with observation of phospho-proteins and cell cycle regulatory protein modifications on transcription. We have markers. The best derived network model rediscovered known increased the throughput of this approach by extensive robotic interactions and contained interesting predictions. Possible automation and have improved the accuracy and precision of applications include the discovery of regulatory interactions, the microarray technology in various ways, using external control design of targeted combination therapies, and the engineering calibration standards to assess improvements. The results of of molecular biological networks. (Models from experiments: expression-profiling mutants in protein kinases, components of combinatorial drug perturbations of cancer cells, S Nelander, WQ the ubiquitin system as well as several global transcription factor Wang, B Nilsson, QB She, C Pratilas, N Rosen, P Gennemark, C complexes will be presented. Sander, MSB, in press). P-04 P-02 Global mapping of the yeast genetic interaction network Cryoelectron tomography: From molecules to systems Boone, Charlie Baumeister, Wolfgang University of Toronto, Banting and Best Department of Medical Max-Planck-Institute of Biochemistry, Martinsried, Germany Research, Toronto, Canada

Electron Tomography (ET) is uniquely suited to obtain 3-D images Objective: To examine yeast genetic interactions quantitatively on of large pleiomorphic structures. While the principles of ET have a global scale, to generate a functional wiring diagram of the cell. been known for decades, its use has gathered momentum only Results: We are examining combinatorial genetic perturbations in recent years. Technological advances have made it possible on a global scale by Synthetic Genetic Array (SGA) analysis. to develop automated data acquisition procedures. This, in turn, We developed a strategy for deriving precise single and double allowed to reduce the total electron dose to levels low enough mutant fitness estimates from SGA genetic interaction screens. for studying radiation sensitive biological materials embedded in We demonstrate the utility of a whole genome catalog of single vitreous ice. As a result, we are now poised to combine the power mutant fitness estimates for assessing the global network of yeast of high-resolution 3-D imaging with the best possible preservation genetic interactions. of the specimen. Conclusions: We show that proper strategies for estimating ET of frozen-hydrated prokaryotic cells or thin eukaryotic cells fitness can yield quantitative genetic interactions on a large-scale. provides 3-D images of macromolecular structures unperturbed Clustering of quantitative genetic interaction profiles defines gene and in their functional environment at molecular resolution (2-4 and pathway function, complementing the information derived nm). Such tomograms contain vast amounts of information; from protein-protein interaction maps. The resultant network essentially they are 3-D images of the cell’s proteome and they provides global view of the pathways that buffer one another and should ultimately enable us to map the spatial relationships of a template for studying the conservation of genetic interactions. macromolecules in a cellular context. However, it is no trivial task to retrieve this information because of the poor signal-to- P-05 noise ratio of such tomograms and the crowded nature of the cytoplasm. Advanced pattern recognition methods are needed for Casting a net for kinases - systematic discovery of In Vivo detecting and identifying specific macromolecules based on their phosphorylation networks structural signature. Provided that high- or medium-resolution Linding, Rune1; Juhl Jensen, Lars2; Ostheimer, Gerry3; Bork, structures of the molecules of interest are available, they can Peer2; Yaffe, Mike3; Pawson, Tony4 be used as templates for a systematic interrogation of the 1ICR, Network & Systems Biology Team, London, United tomograms. Once the challenges of obtaining sufficiently good Kingdom; 2EMBL, Heidelberg, Germany; 3MIT, Boston, United resolution and comprehensive libraries of template structures States; 4SLRI, Toronto, Canada become available, we will be able to map the supramolecular landscape of cells systematically. Directionality in protein signalling networks is due to modulated protein-protein interactions and is fundamental for proper signal progression and response to external and internal cues. This property is in part enabled by linear motifs embedding post- translational modification sites. These serve as recognition sites, guiding phosphorylation by kinases and subsequent binding of modular domains (e.g. SH2 and BRCT). Characterisation

6 ICSB 2008 of such modification-modulated interactions on a proteome- our understanding of how intracellular signals are generated by wide scale requires extensive computational and experimental stimulatory cues, an exceptionally difficult challenge at the present analysis. Thus, we recently developed an approach (NetworKIN) time is to understand how these signals operate in integrated that augments motif-based predictions with the network manner to govern cell behavioral responses. We are undertaking context of kinases and phosphoproteins. The latter provides efforts to address this question by means of a combination of 60%-80% of the computational capability to assign in vivo quantitative, dynamic protein-centric experimental manipulations substrate specificity. NetworKIN pinpoints kinases responsible and measurements with a spectrum of computational mining for specific phosphorylations and yields a 2.5-fold improvement and modeling approaches. Particular application problems of our in the accuracy with which phosphorylation networks can be interest include cell migration, proliferation, differentiation, and constructed. Applying this approach to DNA damage signaling, death, with an emphasis on ascertaining how effectiveness of we showed that 53BP1 and Rad50 are phosphorylated by prospective therapeutics might be predicted. This talk will present CDK1 and ATM, respectively. We describe a scalable strategy to an overview of our perspective and approach, illustrated by an evaluate predictions, which suggests that BCLAF1 is a GSK-3 example vignette in a particular application area.

substrate. In my talk I will review our latest advances in methods Plenary for unravelling phosphorylation mediated cellular interaction P-08 Sessions networks. In particular I will discuss how the combination of quantitative mass-spectrometric technologies and computational Mapping functional elements, regulatory circuits and algorithms (NetworKIN & NetPhorest) together are enhancing variation in yeast and humans mapping of these largely uncharted dynamic networks. By Snyder, Michael combining quantitative measurements of phosphorylation events Yale University, MCDB, New Haven, United States with computational approaches I will discuss how systems level models will help to decipher complex diseases through the ability We have been used microarray and sequenced based methods to predict cellular systems trajectories. to characterize the transcriptome, regulatory elements and structural variation in yeast and humans. These have been used P-06 to assemble regulatory networks, as well as to understand network organization and regulatory evolution. Cracking the regulatory code: Predicting expression patterns from DNA sequence P-09 Segal, Eran Weizmann Institute, Rehovot, Israel Drug interactions and resistance Kishony, Roy Precise control of gene expression lies at the heart of nearly all Harvard Medical School, Boston, United States biological processes. However, despite enormous advances in understanding this process from both experimental and The emergence and spread of antibiotic resistance is focusing theoretical perspectives, we are still missing a quantitative increasing attention to drug-combination therapy. While multi- description of the underlying transcriptional control mechanisms, drug combinations have been studied extensively with respect to and the remaining questions, such as how regulatory sequence their immediate inhibition of bacterial growth, very little is known elements ‘compute’ expression from the inputs they receive, are about their impact on the long-term evolution of drug resistance. still very basic. In particular, it is unclear how evolution of resistance is affected by In this talk, I will present our progress towards the ultimate goal the nature of the interactions—synergy or antagonism—between of developing integrated quantitative models for transcription the drugs. Our combined theoretical experimental approach show regulation, spanning all aspects of the process, including the DNA that antagonism, which is normally avoided in clinical setting, may sequence, regulators, and expression patterns. I will first describe have an advantage in slowing and even reversing the evolution a novel thermodynamic model that computes expression patterns of resistance. Direct competition between antibiotic resistant as a function of cis-regulatory sequence and the binding site and sensitive bacterial cells demonstrate that hyper-antagonistic preferences and expression of participating transcription factors. I drug interactions, in which one drug suppresses the effect of the will show that when applied to the segmentation gene network of other, can invert selection for resistance - mutants that acquire Drosophila, the model accurately predicts the expression of many resistance to one of the drugs lose in competition with their known cis-regulatory modules, even across species, and reveals non-resistant relatives. Further, the range of drug concentrations important organizing principles of transcriptional regulation in the selective for resistance is decreased by antagonistic drug network: that both strong and large numbers of weaker binding interactions. In contrast, synergistic drug pairs, which are normally sites contribute, leading to high occupancy of the module DNA, favored in clinical settings, can sometimes accelerate the rate of and conferring robustness against mutation; and that clustering of evolution. These results point to a tradeoff between the immediate weaker sites permits cooperative binding, which is necessary to efficacy of drug combinations and the future propensity for sharpen the patterns. resistance.

Plenary session 2: Understanding and P-10 curing diseases The cellular uptake of pharmaceutical drugs: A problem not of biophysics but of systems biology P-07 Kell, Douglas; Dobson, Paul The University of Manchester, School of Chemistry and MIB, A quantitative systems approach to understanding how Manchester, United Kingdom signaling networks govern cell behavior Lauffenburger, Douglas Objective.. To understand the cellular uptake of pharmaceutical Massachusetts Institute of Technology, Biological Engineering, drugs. Cambridge MA, United States Results. The enormous attrition rates in the pharmaceutical industry (>90%) are mainly due to lack of efficacy (drugs do Cell behavioral functions are governed by biomolecular not go where you want them to in vivo) and of toxicity (they networks that translate stimulatory cues (e.g., ligand/receptor accumulate where you do not want them to and thereby binding interactions, mechanical stresses, pathogen infection, exert unwanted effects). Both are a consequence of our and other environmental insults) into intracellular signals that poor understanding of how drug molecules enter cells. The regulate transcriptional and post-transcriptional, metabolic, and commonest view, reflected in the famous ‘Rule of 5’ of Chris cytoskeletal processes that effect proximal and ultimate cell Lipinski, is that drugs can partition into the bilayer portion of cell responses. While there is a growing body of work enhancing membranes with rates and extents reflecting their octanol:water

ICSB 2008 7 partition coefficients (log P or log Ko/w), and thereby diffuse P-13 into and out of the cell interior. This is then a biophysical view. Alternatively, and this is considered an ‘exception’, such drugs Genetic and environmental control of yeast metabolism might ‘hitchhike’ on one of the many hundreds of natural SLC Sauer, Uwe; Fendt, Sarah-Maria; Heux, Stephanie; Ewald, carriers or transporters normally used for transporting nutrients Jennifer; Zamboni, Nicola and endogenous metabolites. Even though drugs can bind to Institute of Molecular Systems Biology, ETH Zurich, Zurich, carriers using the same kinds of biophysical forces that can lead Switzerland them to enter and cross phospholipid bilayers, this would then be a mechanistic or systems biology view. In a recent review [1] we To understand how complex networks respond to genetic found some 400 of these ‘exceptions’, including of significantly or environmental challenges is one of the key challenges of Sessions Plenary lipophilic charged molecules, that together with other evidence systems biology. Great strides have been made in our ability suggest that carrier-mediated uptake is the rule and not the to monitor transcriptional and proteome responses, but it is exception. not trivial to predict the functional network response from such Conclusion. This extensive activity of carriers can serve to data. Within metabolism, recent technological advances in mass explain the accumulation of drugs in unexpected places (drugs do spectrometry-based metabolomics and 13C-based flux analysis not simply ‘leak out’) as well as the ability of certain kinds of pro- methods provide us with experimental methods to assess the drugs to enhance uptake. This analysis has profound implications functional state and output of enzyme reaction networks. Using for the design of pharmaceutical drugs, and for the incorporation our recently developed mini-scale 13C-flux and quantitative of this kind of mechanistic information into systems biology metabolomics methods, we quantify here global functional models. responses of yeast metabolism. In particular we focus on i) unravelling the active network of transcriptional regulation that Plenary session 3: Cell regulation actually controls the distribution of metabolic fluxes under different environmental conditions, and ii) on large-scale analysis of how small molecule drugs affect metabolism. P-11 P-14 Single-molecule approach to molecular biology in living bacterial cells Irreversibility of mitotic exit in budding yeast Xie, XiaoLiang Sunney Novak, Bela1; Lopez-Aviles, Sandra2; Kapuy, Orsolya1; Uhlmann, Harvard University, Chemistry and Chemical Biology, Cambridge, Frank2 United States 1University of Oxford, Centre for Integrative Systems Biology, Oxford, United Kingdom; 2Cancer Research UK, LRI, Recent developments on fluorescent proteins and microscopy Chromosome Segregation Laboratory, London, United Kingdom techniques have allowed the probing of single molecules in a living bacterial cell with millisecond time resolution and nanometer Objective: The eukaryotic cell cycle is a simple, cyclic spatial precision. These single-molecule experiments have developmental process by which cells are undergoing through brought new insights into the mechanisms of many processes ordered events of DNA replication (S phase) and chromosome in molecular biology, such as DNA-protein interactions, gene segregation (M phase). The unidirectional progression through regulation, transcription, translation, and replication. I will review the cell cycle is enforced by irreversible transitions: G1/S, the key methods of single-molecule detection and highlight G2/M, mitotic exit etc. These transitions are characterised by numerous examples to illustrate how these experiments are controlled proteolysis of certain cell cycle regulators. For example, contributing to the quantitative understanding of the fundamental cyclin-dependent kinase (Cdk) inhibitors are degraded at G1/S processes of a living cell. transition, while mitotic cyclins are targeted for destruction during mitotic exit. Therefore it is common to explain these irreversible P-12 cell cycle transitions by irreversible degradation of these regulatory proteins. From topology to dynamics and function: Coupled Results: By using budding yeast and computational modeling of feedback loops in signaling networks the cell cycle, we demonstrate that mitotic cyclin degradation by Cho, Kwang-Hyun itself does not render mitotic exit irreversible. Mitotic exit becomes Korea Advanced Institute of Science and Technology (KAIST), irreversible when mitotic Cdk activity drops below a certain Department of Bio and Brain Engineering, Daejeon, Republic of threshold. At this point the positive feedbacks responsible for Cdk Korea activation are disengaged, while the double negative feedback loops are turned to the side of Cdk inhibitors. Objective: Different signal transduction pathways often share Conclusions: These experiments confirm the notion (Novák information through cross-talks. Such cross-talks can form et al., 2007) that irreversible cell cycle transitions are the feedback loops that play important roles in the regulation of cell consequence of systems-level feedback signals, rather than growth, proliferation, and differentiation in response to external protein degradation per se. stimuli. Intriguingly, those feedback loops are frequently found as Reference: Novák, B., Tyson, J.J., Gy~rffy, B. & Csikász-Nagy, a coupled structure in complex cellular circuits. A. (2007): Irreversible cell cycle transitions due to systems-level Results: We have investigated the coupled feedback loops in feedback. Nature Cell Biology 9: 724-728. various cellular circuits and determined the dynamical role of each coupled structure. In particular, we have found that the coupled positive feedbacks enhance signal amplification and bistable characteristics; the coupled negative feedbacks promote homeostasis; and the positive and negative feedbacks together enable a reliable decision making process by properly modulating signal responses and noises. We have further investigated the role of those coupled feedback loops in large-scale networks and evaluated their critical implications in robustness, fragility, and essentiality. Conclusions: In this presentation, I will describe that studying the complex dynamics of multiple feedback loops is a key to understanding the regulatory mechanisms in signal transduction pathways.

8 ICSB 2008 Plenary session 4: From cell to organ to direct target genes or effector molecules are known. Moreover, organism the temporal relationship between the presence of key regulators, the expression of their target genes and the subsequent cell fate determination remains largely unstudied. P-15 Our work attempts to bridge this gap, by combining genetic, genomic and computational approaches to understand Composite kinetics of in vivo assembly and functioning of the transcriptional network that drives the selection of cell multi-protein complexes controlling genome function fates within the mesoderm. By combining ChIP-on-chip Van Driel, Roel1; Luijsterburg, Martijn2; Hoefer, Thomas3; Von through a time-course of Drosophila development we are Bornstaedt, Gesa3; Politi, Antonio4; Moné, Martijn5; Houtsmuller, systematically identifying cis-regulatory module occupancy Adriaan6; Vermeulen, Wim6 during developmental progression. These data are enriched by 1Netherlands Inst. for Systems Biology, Amsterdam, Netherlands; expression profiling of mutant embryos for each transcription 2University of Amsterdam, Swammerdam Inst. for Life Sciences, factor. The topology of the network was unexpected, showing 3

Amsterdam, Netherlands; DKFZ, Heidelberg, Germany; extensive combinatorial regulation and temporal enhancer Plenary 4University of Auckland, Auckland, New Zealand; 5Vrije occupancy. Current work is focused on understanding how these Sessions Universiteit, Amsterdam, Netherlands; 6Erasmus Medical Center, diverse combinatorial binding ‘codes’ give rise to specific patterns Rotterdam, Netherlands of enhancer expression.

Objective: Essentially all processes that control the functioning P-18 of our genome are each carried out by tens of proteins that cooperate in time and space. Examples are transcription initiation Developments in computational physiology complexes, replication machines and DNA repair mechanisms. Hunter, Peter Remarkably little in known about how such complex systems University of Auckland, Bioengineering Institute, Auckland, New function in the living cell. Therefore we have carried out a detailed Zealand analysis of the kinetics of the nucleotide excision repair (NER) system in living cells, which is responsible for the removal of UV- The Physiome Project of the International Union of Physiological induced DNA lesions. Sciences (IUPS) is attempting to provide a comprehensive Results: Results have been integrated in a comprehensive, framework for modelling the human body using computational quantitative and predictive model, which describes the kinetics methods which can incorporate the biochemistry, biophysics of DNA repair in terms of the joint action of eight proteins and and anatomy of cells, tissues and organs. A major goal of the involves six enzymatic steps. project is to use computational modelling to analyse integrative Conclusions: The model gives novel and detailed insight into biological function in terms of underlying structure and molecular DNA repair complex assembly and functioning. Among others mechanisms. A newly formed EU Network of Excellence for the we show that proteins in chromatin-associated complexes Virtual Physiological Human (VPH) is also contributing and, in rapidly exchange with soluble pools in the nucleus, making the particular, addressing clinical applications of the project. formation of a functional multi-protein complex a time-consuming To facilitate model reuse among researchers in computational process. Kinetic proofreading seems essential to drive the repair physiology, two XML markup languages for encoding biological process. Since NER is a paradigm for other chromatin-associated models, CellML (www.cellml.org) & FieldML (www.fieldml.org), are processes, results can be extended to other systems in the being developed. CellML deals with models of so-called ‘lumped nucleus. parameter’ systems, where spatial effects are averaged, and typically involves systems of ordinary differential equations and P-16 algebraic equations. FieldML addresses the spatial variations in cell or tissue properties where the models typically rely on partial The asymmetric dance of the microtubules at the end of differential equations. The two standards can be used together. mitosis These languages, which define the structure of a model, the Nedelec, Francois1; Kozlowski, Cleopatra2; Srayko, Martin3 mathematical equations and the associated metadata, enable (i) 1EMBL, Cell Biology, Heidelberg, Germany; 2EMBL, Heidelberg, automated checking to ensure consistency of physical units used Germany; 3MPI-CBG, Dresden, Germany in the model equations, (ii) models developed by different groups to be combined using commonly agreed ontological terms within Spindle motions observed at anaphase in the first cell stage the metadata, (iii) models to be modularized and used in libraries C.elegans embryo are essential for proper development, because to make it easier to create complex models by importing simpler they eventually place the mitotic spindle on the posterior side, ones. and in this way establish an asymmetric division. Using confocal Model repositories based on these standards and implementing a microscopy, we observed microtubules near the cell cortex during wide variety of models from peer-reviewed publications have been this process, which indicated how pulling forces acting on the developed (www.cellml.org/models) and open source software spindle might be produced. From these observations, a physical tools for creating, visualizing and executing these models are model was built, which showed that such force production currently available (www.cellml.org/tools) and under continuous mechanism is crucially determined by the assembly/disassembly development. dynamics of fibers and their mechanical properties. Interestingly, The application of this framework to modeling the heart and other this model proves that posterior displacement arises robustly from organs will also be presented. the physical properties of the system. Plenary session 5: New approaches to P-17 biotechnology Gene regulatory networks during development: Dissecting the logic P-19 Furlong, Eileen Computational modeling and simulation lead to the EMBL, Heidelberg, Germany development of MM-121, a human monoclonal antibody ErbB3 antagonist One of the major challenges in developmental biology is to Schoeberl, Birgit; Pace, E. A.; Fitzgerald, J. B.; Harms, B.D.; understand how pluripotential fields of cells become specified West, K. A.; Kumar, A.; Kudla, A. J.; Nielsen, U.B. and organised to form complex tissues. Genetic studies have Network Biology, Cambridge, United Kingdom identified a number of essential transcription factors required for cell fate specification, however little is known about the molecular The epidermal growth factor receptor and ErbB2 have been mechanisms by which these regulators function. Few of their heavily targeted by therapeutic agents, with some success.

ICSB 2008 9 Understanding the relative importance of the ErbB receptors analysis during growth on different carbon sources, e.g. glucose, for inducing activation of downstream signaling cascades is xylose, xylan, starch and arabinan. For transcription analysis complicated for a number of reasons: the ErbB receptors homo we are using the Affymetrix platform and for A. niger we use a and heterodimerize to become activated, are expressed to custom-designed DNA array, that covers all of the about 12,000 differing degrees in different cancer types, and undergo ligand- ORFs identified in the recently released DOE sequencing effort. specific trafficking. Hence, we have taken a systems approach to In the lecture there will be given an overview of the application the problem and developed a computational model of the ErbB of these two fungi as cell factories, the advancement towards a signaling network for the purpose of developing more efficacious wider use of these organisms and some projections for the future ErbB targeted therapeutics. chemical production using fermentation will be given. The computational model describes the dynamics of all four ErbB Sessions Plenary receptors in response to ligands that bind to EGFR (Betacellulin) P-21 or ErbB3 (Heregulin). Ligand-induced dimerization, receptor internalization, degradation and downstream signaling of the Evolvability and hierarchy in rewired bacterial gene PI3K-Akt cascade are included. To train the model the activation networks status of all ErbB receptors and pAkt, an important regulator of Serrano, Luis cell proliferation and survival, were measured after stimulation CRG-EMBL Systems Biology Unit, Barcelona, Spain with each ligand in ADRr cells: 12 time points ranging up to 2 hours post stimulation and 8 ligand concentrations spanning Sequencing DNA from several organisms has revealed that three orders of magnitude. This dataset constrained the dynamic duplication and drift of existing genes has primarily molded the behavior of the model for parameter estimation and the model contents of a given genome. was then tested using DU145 and ACHN cell lines. Though the effect of knocking out or over-expressing a Sensitivity analysis was performed of the system to determine particular gene has been studied in many organisms, no the key proteins in the network for activating pAkt over a diverse study has systematically explored the effect of adding new range of heterogeneous ErbB receptor profiles. ErbB3 was links in a biological network. To explore network evolvability, found to be an ultra-sensitive inhibition point in response to we constructed 598 recombinations of promoters (including either Betacellulin or Heregulin. This prediction was verified by regulatory regions) with different transcription or ó-factors in developing and testing a novel anti-ErbB3 antibody, MM121, in Escherichia coli, over the genetic background of the wild-type. cell based assays. MM-121 was also found to be more effective Here we show that ~95% of new networks are tolerated by the at inhibiting Heregulin-induced pErbB3 and pAkt than other bacterial cell, that very few alter growth, and that expression therapeutic agents targeting the ErbB pathway further supporting levels correlate with the position of the factor in the wild-type the findings. network hierarchy. Most importantly, we find that certain networks The kinetic parameters of MM-121 were also used to build an in consistently survive over the wild-type under various selection silico version of the inhibitor. The model was then used to predict pressures. Therefore new links in the network are rarely a barrier the IC50 values of pAkt and pErbB3 in response to MM121 in for evolution and can even confer a fitness advantage. a number of cell lines stimulated with Heregulin or Betacellulin and these predictions were experimentally verified. These in vitro P-22 findings were further validated in vivo using multiple xenograft models. Why not edit DNA the way we edit text? In summary, we describe 1) how a systems approach to drug Shapiro, Ehud1; Kaplan, Shai2; Shabi, Uri3; Ben-Yehezkel, Tuval2; discovery has suggested a potentially better mechanism of Linzhiz, Gregory2 inhibiting the ErbB pathway and 2) the ability of our computational 1Weizmann Institute of Science, Computer Science and Biological model to accurately simulate the effect of an anti-ErbB3 Chemistry, Rehovot, Israel; 2Weizmann Institute of Science, monoclonal antibody on signaling in a variety of cancer cell type. Biological Chemistry, Rehovot, Israel; 3Weizmann Institute of Science, Computer Science and Applied Math, Rehovot, Israel P-20 Polymerase Chain Reaction (PCR) is the DNA-equivalent of Industrial systems biology: Yeast and filamentous fungi as Gutenberg’s movable type printing, both allowing large-scale cell factories for sustainable production of chemicals automated replication of a piece of text. DNA synthesis machines Nielsen, Jens are the DNA-equivalent of mechanical typesetting machines, Chalmers University of Technology, Department of Chemical and both ease the setting of a piece of text to be replicated. What is Biological Engineering, Göteborg, Sweden the DNA-equivalent of the word processor, which allows easy composition and editing of a piece of text? Here we present The yeast Saccharomyces cerevisiae and the filamentous the DNA-equivalent of a word processor, a framework for DNA fungus Aspergillus niger are used extensively in the fermentation processing that supports all basic editing operations on DNA industry for the production of a range of different products, such molecules including insert, delete, replace, cut&paste and as fuels, fine chemicals, food ingredients, enzymes, protein copy&paste. The framework incorporates our earlier work on drugs, vaccines, beer, wine and bread. In connection with recursive composition and error-correction of DNA molecules further development of bioprocesses for sustainable production and offers a general, efficient and automatable method for DNA of fuels and chemicals these fungi are interesting versatile cell editing, synthesis, and variant-library construction. factories. S. cerevisiae is interesting as it is easy to perform directed genetic modifications and an extensive systems biology Plenary session 6: Genetic variation and tool box is available for identification of metabolic engineering targets. A. niger is interesting as it tolerates low pH, can utilize a evolution wide range of carbon sources and has relatively high conversion rates. In connection with developing S. cerevisiae and A. niger P-23 as general cell factory platforms we have established a number of systems biology technologies that can be used for enhancing Causally cohesive genotype-phenotype models - systems the metabolic engineering of these organisms in the future. These biology meets genetics techniques include genome scale metabolic models, in which Omholt, Stig W. there is a direct link between reactions, enzymes and genes and Norwegian University of Life Sciences, Centre for Integrative hereby it is possible to rapidly map transcriptional changes on Genetics (CIGENE), Aas, Norway global metabolic maps. For A. niger the model contains a large number of extracellular reactions for the degradation of complex A comprehensive understanding of how genetic variation causes carbohydrates, and these reactions form a complex network phenotypic variation of a complex trait is a long-term disciplinary which has been analyzed using genome-wide transcription goal of genetics. Unless we succeed in making mathematical

10 ICSB 2008 conceptualizations demonstrating how genetic variation becomes P-26 manifested in phenotypic variation at various systemic layers up to the whole-organism level, we cannot pretend to have a real Genetic diversity in gene expression stochasticity and quantitative genetics theory. Compared to the broader class of chromatin epigenetic states systems biology models, these models can be distinguished by Yvert, Gael1; Ansel, Juliet1; Nagarajan, M.1; Veyrieras, Jean- having an articulated relation to the individual’s genotype. They Baptiste1; Bottin, Hélène1; Damon, Christelle1; Steinmetz, Lars2; seek to account for the effects of genetic variation through a De Dieuleveult, Maud1; Fehrmann, Steffen1 description of the proximal processes linking the two domains 1Ecole Normale Superieure de Lyon, LBMC/CNRS, LYON, France; in terms of regulatory principles and mechanisms such that 2EMBL, Heidelberg, Germany phenotypic values are emergent properties of lower-level processes. We call these types of models causally cohesive Gene expression has been shown to vary greatly among natural genotype phenotype models (or cGP models for short) as they populations and between congenic lines of model organisms. at some given level of resolution have the quality of causing Covering yeast, human, worm, plants or mouse, many past

components involved in a genotype-phenotype relation to cohere studies showed that the expression of most genes is subjected Plenary in a logically consistent and ordered way. A state-of-the-art to heritable variation with a complex genetic architecture. All were Sessions description of how cGP models have been, and can be, used based on samples where gene products are estimated from many to understand phenotypic variation within evolutionary biology, (>millions) cells and thus reflected average signatures. In the production biology and biomedicine will be provided. simple context of the yeast S. cerevisiae, we examined in more details two aspects: the effect of genetic variation on single-cell P-24 expression levels, and the diversity of chromatin epigenetic states among wild isolates. On the relationship between robustness and evolvability Using a simple inducible GFP construct integrated in the Wagner, Andreas genome of various yeast isolates, we found that stochasticity University of Zurich, Dept. of Biochemistry, Zurich, Switzerland in gene expression (cell-to-cell variability) varied with genetic backgrounds. Crossing strains with different levels of Mutational robustness and evolvability, a system´s ability stochasticity, we observed that cell-to-cell variability was to produce heritable phenotypic variation and, ultimately, highly heritable and under a complex genetic control. We evolutionary innovations, harbour a paradoxical tension. On mapped Quantitative Trait Loci (QTL) of noise which led to one hand, high robustness implies low evolvability. On the other the demonstration that cell-to-cell variability increased when hand, both experimental and computational analyses of neutral transcriptional elongation was impaired. This model study shows networks indicate that robustness enhances evolvability. I here that genetic variation can have probabilistic effects, and we resolve this tension using RNA genotypes and their secondary suggest that cases of incomplete penetrance in multicellular structure phenotypes as a study system. To resolve the tension, organisms may result from such effects. one must distinguish between robustness of a genotype and a We also compared the distribution of nucleosomes and histone- phenotype. I confirm that genotype (sequence) robustness and modifications along the genome of unrelated wild strains using evolvability share an antagonistic relationship. In stark contrast, high-resolution mapping. An estimate of natural epigenetic phenotype (structure) robustness promotes structure evolvability. diversity and its correlation to gene expression will be presented. A consequence is that finite populations of sequences with a robust phenotype can access large amounts of phenotypic Plenary session 7: Future challenges variation while spreading through a neutral network. Population- level processes and phenotypes rather than individual sequences are key to understand the relationship between robustness P-27 and evolvability. My observations may apply to other genetic systems where many connected genotypes produce the same Genomics, computation, and the nature of biological phenotypes. understanding Botstein, David P-25 Lewis-Sigler Institute, Princeton University, Princeton, New Jersey, United States Noise, penetrance and cell fate in bacterial development Elowitz, Michael B. Genomic sequences provide direct evidence that the basic CalTech, Pasadena CA,, United States cellular functions of all organisms are carried out by genes and proteins whose primary sequences are simply related by Development normally proceeds through a consistent sequence evolutionary descent. The genome sequences also allow us, of events that occurs similarly in all wild-type individuals. However, for the first time, to study all the genes of a single organism developmental mutations often exhibit partial penetrance, simultaneously. We can now use high-throughput methods affecting the fate of individual organisms differently, even within an (e.g. DNA microarrays to study patterns of gene expression isogenic population in a homogeneous environment. Here, using and genome rearrangements or mass spectrometry to study Bacillus subtilis sporulation as a model developmental system, metabolite dynamics). These system-level experiments, and the I will discuss a stochastic cell fate determination network that new computational and statistical methods required to analyze operates when inter- compartmental signaling components are them, allow us to begin to understand not only the functions of mutated. Competition between molecular processes results in a individual genes and proteins, but also how these work together set of discrete alternative fates not found in the wild-type strain, as interacting systems, in the settings of both normal growth and including the ability to form two “twin” spores, rather than one, disease. in a single mother cell. This system thus enables us to analyze the inter-related roles of noise, cell fate determination, and partial P-28 penetrance in bacterial development and its evolution. Systems biology for global climate and energy solutions Kitano, Hiroaki Systems Biology Institute, Tokyo, Japan

There is a clear and present danger in this planet caused by global climagte change and energy/resource issues. It is about fragilty of our civilization that are built to depend on high rate of resource use to maintain high comsumption life style. Our civilization today is adapted to stable climate, abandancy

ICSB 2008 11 of resources such as oil, water, rare metals, and terrestrial capacity to absorbe pollutants. These assumptions are now seriously questioned. As Jared Diamond indicated in his book “Collapse”, collapse of civilizations in the past often associated with destruction of local ecological system. The problem today is ecological systems are being perturbed at global scale which implies potential risk that collapse may be at the global scale, rather than a locally contained event. Now, the question is what systems biology can do to avoid catastrophy. Earth is a complex systems of geosphere, biosphere, Sessions Plenary and econosphere. Lovelock and Marglis conceptualized this as GAIA, a hypothetical life form at the terristerial scale. Understanding sysems biology of GAIA, and series of specific research to understand the issues and development of coutermeasures are essential. Biofuels from non-agricultural products, water quality improvement, recycling of resources from waste by biological means, and preservation and restration of coral reefs and tropical rainforst are some of topics that are often discussed. Progress in such issues require system-oriented approaches at multiple scale. Systems biology should be able to contribute to this high priority issue.

P-29

Systems medicine, transformational technologies and the emergence of P4 medicine Hood, Lee Institute of Systems Biology, Seattle, United States

The challenge for biology in the 21st century is the need to deal with its incredible complexity. One powerful way to think of biology is to view it as an informational science. This view leads to the conclusion that biological information is captured, mined, integrated and finally executed by biological networks. Hence the challenge in understanding biological complexity is that of deciphering the operation of dynamic biological networks across the three time scales of life -evolution, development and physiological responses. Systems approaches to biology are focused on delineating and deciphering dynamic biological networks and their interactions with simple and complex molecular machines. I will focus on our efforts at a systems approach to disease - looking at prion disease and cancer. I will also discuss the emerging technologies (measurement and visualization) that will transform medicine over the next 10 years. It appears that systems medicine, together with pioneering changes in DNA sequencing and blood protein measurements (nanotechnology) and as well as the development of powerful new computational and mathematical tools will transform medicine over the next 5-20 years from its currently reactive state to a mode that is predictive, personalized, preventive and participatory (P4). This will in turn lead to the digitalization of medicine - with ultimately a profound decrease in the cost of healthcare.

12 ICSB 2008 Dedicated Orals Orals Dedicated Dedicated session 1-1: Cell-regulation – for its automated identification from high-throughput UPLC/ metabolism MS-based lipidomics experiments. The utility of our approach is demonstrated by its application to the lipidomic characterization of the fatty liver in insulin resistant mice. We investigate the DS1-1-01 changes of correlation structure of the lipidome using multivariate analysis, as well as reconstruct the pathways for specific Predicting flux control distributions in metabolic networks molecular species of interest using available lipidomic and gene for which limited kinetic information is available expression data. Snoep, Jacky1; Conradie, Riaan1; du Preez, Franco1; van Gend, Conclusion: Our methodology facilitates identification and Carel1; Bruggeman, Frank2; Westerhoff, Hans2 interpretation of high-throughput lipidomics data. In the context 1University of Stellenbosch, Biochemistry, Stellenbosch, South of the ob/ob mouse liver profiling, we have identified the parallel Africa; 2Vrije Universiteit, Molecular Cell Physiology, Amsterdam, associations between the elevated triacylglycerol levels and the Netherlands ceramides, as well as the putative activated ceramide-synthesis pathways. Objective: Rational approaches for metabolic engineering of Dedicated biochemical networks are hampered by the limited availability of DS1-1-03

Orals kinetic information for the enzymes catalyzing the reactions in the system. We have developed strategies that do not need detailed Limited number of transcription factors control metabolic kinetic information but can still be used to pinpoint enzymes that fluxes in yeast have a high flux control. The first of the strategies considers the Fendt, Sarah-Maria; Sauer, Uwe importance of the binding and thermodynamic terms in enzyme ETH Zurich, Zurich, Switzerland kinetic rate equations and allows the prediction of an enzyme’s importance in flux control on the basis of the mass-action ratio Objective: Biological processes are transcriptionally, post- and the equilibrium constant for the reaction it catalyses. The transcriptionally and allosterically regulated but the impact varies. second strategy considers the importance of flux scaling in In this study we investigate the influence and thus the control of biochemical networks and can be applied to branched pathways. transcription factors (TF) on metabolic fluxes. To elucidate the For this analysis the sensitivities of enzymes for the branching network of TF’s that control metabolic fluxes we analyzed 119 metabolite are important to calculate flux control distributions. haploid S. cerevisiae mutants each lacking a TF by large scale Both the analysis methods have been derived analytically for 13C flux analysis. These are 70 % of all TF‘s related to metabolism core models and were tested numerically on large numbers of in and stress response and they were tested under different growth silico networks, with random and scale free structures. In addition conditions, including two carbon sources and various stresses. the methods were applied to the metabolic models (showing Results: We identify 5 new flux controlling interactions of TF’s, steady state behavior) that are available in the JWS Online and provide quantitative and condition-specific information for 13 TF‘s Biomodels databases. that were generally expected to have an impact on fluxes and Results: A good correlation was observed between the degree confirm 5 known interactions. In total we find 23 TF‘s that control that a reaction is away from equilibrium and its flux control metabolic fluxes under at least one of the 5 tested conditions. coefficient. The extent that the total pathway is away from All of them have an impact on the respiratory tricarboxylic-acid equilibrium effects this correlation. The correlation could for simple (TCA) cycle flux. Two of the 23 control additionally the glycolytic systems be derived analytically and was shown numerically for flux when glucose is the carbon source. No other metabolic models ranging from core models to detailed kinetic models of fluxes where influenced by the tested TF’s under the 5 conditions. metabolic pathways. The second approach has shown that in Eleven TF’s are controlling the respiratory TCA cycle under branched pathways where the branches have fluxes that differ only one condition. Four TF’s are identified as repressors of the several orders of magnitude, a simplified analysis on the basis respiratory TCA cycle flux in all conditions where glucose is the of sensitivity for the branching metabolite could be used to carbon source independently from the applied stresses. Eight TF’s determine whether the main flux control resides in the branch or are crucial to activate the respiratory TCA cycle on glucose and in the main pathway. on galactose. Conclusions: The study shows that with limited kinetic Conclusions: The data show that TF’s are important for the information important guidelines for control distribution in control of the respiratory TCA cycle. Nevertheless, a topological metabolic networks can be formulated. interaction network based on a binding and expression database identify 42 interactions of transcription factors with respiratory DS1-1-02 enzymes and 54 with glycolytic enzymes. Our data however reveal a much smaller network of actually flux controlling TF‘s. Reconstruction of lipid pathways at cellular and systemic level DS1-1-04 Oresic, Matej1; Yetukuri, Laxman1; Vidal-Puig, Antonio2; Medina- Gomez, Gema2; Katajamaa, Mikko1 Thermodynamically-consistent reduced-order modeling of 1Technical Research Centre of Finland, Espoo, Finland; the oxygen response of escherichia coli 2Canbridge University, Cambrudge, United Kingdom Ederer, Michael1; Sauter, Thomas2; Bettenbrock, Katja1; Sawodny, Oliver2; Gilles, Ernst Dieter1 Objective: Lipids are an important and highly diverse class of 1Max Planck Institute for Dynamics of Complex Technical molecules having structural, energy storage and signaling roles. Systems, Magdeburg, Germany; 2Institute for System Dynamics, Modern analytical technologies afford screening of many lipid Universität Stuttgart, Stuttgart, Germany molecular species in parallel. One of the biggest challenges of lipidomics is elucidation of important pathobiological phenomena Objective: The response of Escherichia coli to a varying from the integration of the large amounts of new data becoming oxygen availability is mediated by several genetic and enzymatic available. regulation systems. The adaptation to a change in oxygen Results: We present computational and informatics approaches availability involves a major reorganization of metabolic fluxes and to study lipid molecular profiles in the context of known metabolic concentrations. Thus, a consistent understanding of the oxygen pathways and established pathophysiological responses, utilizing response cannot be achieved by a reductionist’s approach only. information obtained from modern analytical technologies. A computational model of the oxygen response can provide us In order to facilitate identification of lipids, we compute the with a sound understanding of the interaction of the involved scaffold of theoretically possible lipids based on known lipid processes. building blocks such as polar head groups and fatty acids. Results: We built a computational model of the steady- Each compound entry is linked to the available information on state oxygen response of Escherichia coli in a glucose limited lipid pathways and contains the information that can be utilized chemostat. It contains the glycolysis, the tricarboxylic acid cycle,

14 ICSB 2008 the electron transport chain and several fermentative pathways. DS1-1-06 Further, it contains the genetic regulation by the transcription factors ArcA, FNR, PdhR and CRP. In order to guarantee the Modeling the glycolytic pathway in human skeletal muscle thermodynamic feasibility of the model, we use the recently tissue; understanding the regulation of its hundredfold developed Thermodynamic-Kinetic Modeling formalism (Ederer dynamic range & Gilles, Biophys J, 92(6), 2007). To reduce the number of state Schmitz, Joep; Jeneson, Jeroen; Nicolay, Klaas; Hilbers, Peter; variables and parameters we extensively apply rapid-equilibrium van Riel, Natal assumptions to the metabolic reactions that are know to proceed Eindhoven University of Technology, Biomedical Engineering, near equilibrium. The model parameters were adapted to fit Eindhoven, Netherlands experimental data (uptake and excretion fluxes, intracellular nadh/ nad ratio and several mRNA and enzyme activities, Alexeeva et Objective: During vigorous exercise, glycolytic flux in human al., J Bact, 2000, 2002, 2003). skeletal muscle tissue increases more than hundredfold. Although Conclusions: The model predicts intracellular fluxes and the glycolytic pathway in general has been the subject of many concentrations as well as the activity of the transcription factors experimental and computational studies, it is still largely unknown for varying oxygen availability. Further, with the model we can how this large dynamic range of skeletal muscle glycolysis is perform extensive in silico mutant studies. The above introduced achieved and controlled. The hypothesis was tested that the computational model provides a coherent and global picture bottleneck is not our understanding of the high flux regime, but of the oxygen response of E. coli. In the future the SUMO instead in homeostasis of rest metabolism. We augmented the

consortium will refine the model by testing several of the model computational model of glycolysis of Lambeth and Kushmerick Orals predictions on mutant behavior. (Ann Biomed Eng. 2002, 30(6): 808-827). Model simulations were Dedicated Acknowledgements: This work was supported by SysMO; validated with in vivo 31P NMR measurements. project number 3 (Systems Understanding of Microbial Oxygen Results: The existing model with mammalian kinetic parameters Responses, SUMO); www.sysmo.net. ME acknowledges also was re-parameterized for human skeletal muscle by estimating support from the German Bundesministerium für Bildung und Vmax parameters based upon in vivo measurements of maximal Forschung (BMBF, FORSYS initiative). We thank all members of glycolytic fluxes. Noninvasive, in vivo measurements of muscle SUMO for fruitful discussions. energetics were obtained with 31P NMR spectroscopy. 31P NMR spectra were recorded during high intensity, in-magnet bicycle DS1-1-05 exercise performed until exhaustion and dynamics of [PCr], [ATP], [Pi], [Glucose-1P], [Glucose-6P] and [Fructose-6P] were obtained. Plasticity of genetic interactions in yeast metabolism Both the original and the re-parameterized model predicted Papp, Balazs1; Harrison, Richard2; Pal, Csaba1; Oliver, Stephen an unphysiologically high resting flux. Including an additional G.3; Delneri, Daniela4 suppressing regulatory mechanism yielded a model able to 1Biological Research Center, Szeged, Hungary; 2The University reproduce the 31P NMR data. of Manchester, Manchester, United Kingdom; 3University of Conclusion: The re-parameterized, augmented model was able Cambridge, Cambridge, United Kingdom; 4The University of to reproduce the hundredfold dynamic range of glycolytic flux in Manchester, Manchester, United Kingdom human skeletal muscle tissue; model simulations agreed very well with in vivo measurements of muscle metabolism. We speculate Objective: Why are most genes dispensable under laboratory that detachment of enzymes from the cytoskeleton could be an conditions? It has been suggested that the phenotypic impact of important mechanism in the regulation of the glycolytic resting gene deletions may depend on the environment or the presence rate. of compensatory mechanisms (mutational robustness). However, the potential links between environment-specific functions DS1-1-07 and robustness against harmful mutations have remained largely unexplored. It may well be that these theories on gene Genome-wide regulation of γ-linolenic acid biosynthesis in dispensability are not mutually exclusive. Here, we analyse the Saccharomyces cerevisiae interaction between these two forces by exploring the condition Ruenwai, Rawisara1; Laoteng, Kobkul2; Tanticharoen, Morakot3; dependence of synthetic genetic interactions that define Cheevadhanarak, Supapon1; Petranovic, Dina4; Nielsen, Jens4 redundant functions and alternative pathways. 1King Mongkut’s University of Technology Thonburi, School of Results: We performed systems-level flux balance analysis of Bioresources and Technology, Bangkok, Thailand; 2Biochemical the yeast (Saccharomyces cerevisiae) metabolic network to Engineering and Pilot Plant Research and Development Unit, identify synthetic genetic interactions and then tested the model’s Biotec at Kmutt, Bangkok, Thailand; 3National Center for Genetic predictions with in vivo gene deletion studies. Experimental Engineering and Biotechnology, Pathumthani, Thailand; 4Chalmers analyses not only confirmed predictions of the model but also University of Technology, Department of Chemical and Biological showed that investigation of false predictions may both improve Engineering, Gothenburg, Sweden functional annotation within the model and also lead to the discovery of higher-order (i.e. trigenic) genetic interactions. We Objective: Gamma-linolenic acid (GLA), a C18:3 ω6 found that the majority (~85%) of synthetic genetic interactions polyunsaturated fatty acid, is conditionally essential fatty acid are restricted to certain environmental conditions, partly due that is required for maintain the homeostasis of human health. to the lack of compensation under some (but not all) nutrient Regulatory mechanism(s) involved in the GLA biosynthesis in fungi conditions. is presently in its infancy because of unavailability of their genome Conclusions: These findings suggest that compensating gene sequence. Therefore, profiling of transcriptome, fatty acid and pairs have at least partially independent functions, and hence lipid metabolites in the recombinant S. cerevisiae carrying Δ12- compensation is only a by-product of their evolutionary history. and Δ6-desaturase genes of Mucor rouxii were done and the data Our work supports the view that functional redundancy may be were integratively analyzed. more apparent than real, and it offers a unified framework for the Results: Heterologous expression of M. rouxii desaturase genes evolution of environmental adaptation and mutational robustness. in S. cerevisiae was constitutively regulated by the M. rouxii Δ9- desaturase promoter. The results showed a similar pattern of physiological characteristics between the recombinant yeasts and a reference strain containing empty vector. Under glucose- limited cultivations, the transcriptome and metabolite profiling of the recombinant yeasts, carrying M. rouxii Δ12- and Δ6-desaturase genes, were analyzed by comparison with the reference strain. The accumulation of the desaturated products, LA and GLA, was observed in S. cerevisiae bearing both Δ12- and Δ6-desaturase genes. The proportion of GLA, varied between individual lipid

ICSB 2008 15 classes, in which TAG and FFA contained a high proportion way, and to add semantics information using perennial and of GLA. To interpret the results of the transcriptome data in a extendable controlled vocabularies. Results: The Minimal pathway context, the significant expressed genes at an adjusted Information About a Simulation Experiment (MIASE) project is a cut-off value P < 0.01 were painted on the cellular overview community effort that aims at describing the information needed chart by using Pathway Tools Omics Viewer(http://pathway. to repeat a numerical simulation experiment derived from a yeastgenome.org:8555 /expression.html). The result revealed that given set of quantitative models. MIASE guidelines describe: 1) genes involved in the TCA cycle, pentose phosphate pathway, simulation settings (type of simulation and the corresponding amino acid and ergosterol metabolism were significantly changed parameters) 2) model references and information about changes when compared to the reference strain. done to the model before the simulation was run, 3) different Conclusions: Towards efficient pathway manipulation for GLA tasks performed on certain models using certain simulation biosynthesis in heterologous host, changes in gene expression specifications, 4) desired outputs of the simulation (graphs and relevant biological processes, as a result of the genetic or reports, which variables: to plot etc.). The MIASE-OM is a perturbation of the yeast transformants, were identified based formal representation of the requirements defined in the MIASE on genome-wide analysis. This study revealed the physiological guidelines. MIASE-OM is described using UML and is the basis response of the yeast cells by modulating central metabolism and for data models, for example in XML Schema representation. In Dedicated lipid metabolism to maintain cellular homeostasis. order to describe precisely the simulation settings, we developed

Orals KiSAO, an ontology of kinetic simulation algorithms. Preliminary DS1-1-08 versions of MIASE-OM, covering only a subset of simulations used in Systems Biology have already been implemented by a set Mathematical model of low density lipoprotein (LDL) of simulation tools. endocytosis by hepatocytes Conclusion: MIASE and KiSAO complete and extend MIRIAM O’Malley, Brendan1; Wattis, Jonathan2; Blackburn, Hannah3; and SBO in order to fully specify the semantics of quantitative Pickersgill, Laura3; Panovska, Jasmina3; Tindall, Marcus4; models and the associated simulations. Jackson, Kim5 1Unilever R&D Colworth, Sharnbrook, United Kingdom; 2University DS1-2-02 of Nottingham, Centre for Mathematical Medicine, Nottingham, United Kingdom; 3Unilever R&D Colworth, Sharnbrook, United Sharing biological data and knowledge on the semantic Kingdom; 4University of Oxford, Centre for Mathematical Biology, web Oxford, United Kingdom; 5University of Reading, Hugh Sinclair Ruttenberg, Alan Unit of Human Nutrition, Reading, United Kingdom Science Commons, Cambridge, United States

Individuals with elevated levels of plasma low density lipoprotein Objective: The Semantic Web, with its emphasis on (LDL) cholesterol (LDL-C) have an enhanced risk of cardiovascular interoperable representation languages suitable for computational disease. LDL levels are in part controlled by their rate of clearance use, and on easy access via standard web protocols, offers from the plasma. This occurs via a process referred to as a technological basis for building knowledge bases with receptor-mediated endocytosis, which occurs predominately in interdisciplinary coverage. Semantic Web technologies reduce the liver. A series of classical experiments delineated the major the effort involved with building and maintaining such knowledge steps in the endocytotic process; apolipoprotein B-100 present bases, while maximizing access, and therefore increase the on LDL binds to a specific receptor (LDL receptor (LDL-R)) in efficiency of knowledge sharing. In this talk, I will give an overview specialized areas of the cell surface called clatharin-coated pits. of the current state of Semantic Web technologies as applicable The pit comprising the LDL-LDL-R complex is internalized within a to representing the various domains that are the components of cytoplasmic endosome. Fusion of the endosome with a lysosome systems biology. I will focus specifically on OWL and SPARQL, leads to degradation of the LDL into its constituent parts, which the methodology of the OBO Foundry, and discuss a number are released for reuse by the cell. of OBO and OBO Foundry ontologies that form a foundation on We have formulated a mathematical model of LDL endocytosis, which to build. Taken together, the standards and the practices including its regulation by intracellular cholesterol levels, consisting present the first potentially viable approach for enabling data of a system of ordinary differential equations. We validate our integration on the scale of the Web. model against existing in vitro experimental data, and we use it Results: The Neurocommons is a prototype biomedical to explore differences in system behavior when a single bolus knowledge base built on these technologies. Utility of the of extracellular LDL is supplied to cells, compared to when a knowledge base is demonstrated by reviewing a series of continuous supply of LDL particles is available. Whereas the queries that provide compact and interesting answers to former situation is common to in vitro experimental systems, the precise questions relevant to the understanding of disease, latter better reflects the in vivo situation. Together with numerical including some that are not easily answered by existing public simulations we use asymptotic analysis to study the longtime bioinformatics systems. behaviour of model solutions. The implications of model-derived All components of the knowledge base are freely available, insights for experimental interpretation and design are discussed. enabling readers to reconstruct the knowledge base and experiment with this new technology. Dedicated session 1-2: Conclusion: By adopting approaches discussed, repositories might, at the same time as providing for the specific needs of the Standards and repositories communities they serve, also make data and knowledge they encompass available to combine with other repositories large and DS1-2-01 small, enabling powerful query and computational experiments with much less effort than was previously possible. MIASE - The minimum information about a simulation experiment Le Novere, Nicolas1; Koehn, Dagmar2 1EMBL-EBI, Hinxton, United Kingdom; 2University of Rostock, Rostock, Germany

Objective: Effective understanding and reuse of quantitative models necessitate to precise the procedures requested in order to run the adequate simulation experiments and obtain the expected results. In addition to the model structure, a complete description requires to provide a minimal core of agreed-upon information about its usage, to encode it in a computer-edible

16 ICSB 2008 DS1-2-03 by employing a naïve Bayes classifier, and the implementation of SBGN based visualisation. Targeted development of registries of biological parts Blauvelt, Meagan; Cai, Yizhi; Cooper, Krisal L.; Crasta, Oswald; DS1-2-05 DeLalla, Emily C.; Evans, Clive; Folkerts, Otto; Lyons, Blair M.; Mane, S.P.; Shelton, Rebecca; Sweede, Matthew A.; Waldon, Annotating experimental kinetic data for quantitative Sally A.; Peccoud, Jean modelling: The SABIO-RK database Virginia Tech, Virginia Bioinformatics Institute, Blacksburg, United Golebiewski, Martin; Kania, Renate; Krebs, Olga; Mir, Saqib; States Weidemann, Andreas; Wittig, Ulrike; Rojas, Isabel EML Research gGmbH, Heidelberg, Germany Objective The design and construction of novel biological systems by combining basic “building blocks” represents a Objective: Scientific communication in systems biology needs dominant paradigm in Synthetic Biology. Since 2000, the standards, shared vocabularies and unambiguous identifiers standardization of these basic parts has been regarded as crucial to avoid misinterpretations. Standards are especially important to the transition from the ad-hoc methods of traditional genetic when gathering data from different sources like for the set-up of engineering to the industrial-scale process being contemplated large biochemical models. To piece the puzzle of individual results by engineers leading this emerging field. As synthetic constructs together computer systems are needed that unify the information become more complex, it becomes necessary to develop and make it comparable through annotating the data to identifiers

registries of biological parts for various applications, organisms, that are shared across the resources. Orals etc. Results: To provide quantitative experimental data, we have Dedicated Results We have analyzed the abstraction hierarchy of the MIT developed SABIO-RK (http://sabio.villa-bosch.de/SABIORK/), Registry of Standard Biological Parts by looking at inclusion a database system offering comprehensive information about relationships among part sequences in its database. The biochemical reactions and corresponding kinetic data. It not observation that the sequences of some entries were included only describes participating or modifying molecules and kinetic in the sequence of other entries placed lower in the Registry parameters of reactions, but also provides the kinetic rate abstraction hierarchy indicates that Registry users experienced equations and the environmental conditions for parameter difficulties in using the Registry proposed hierarchy. We have also determination. The database is populated by merging information performed a global evaluation of the quality of the DNA repository from other databases with data that we manually extract from associated with the Registry which unraveled widespread literature. It can be accessed manually by a web-interface or discrepancies between the clones published sequences and their programmatically by web services. assembled sequence. The collected data is standardized to a uniform format and Conclusions Guidelines for the management of registries of structure. This comprises the usage and development of biological parts have been proposed as a result of this work. We controlled vocabularies and algorithms to unify the data. recommend a targeted approach that makes a clear distinction Entities and expressions in SABIO-RK are annotated to other between basic parts and designs. Instead of attempting to resources and biological ontologies. The data can be exported develop a monolithic abstraction hierarchy and part categorization in SBML (Systems Biology Markup Language) together with the scheme, we show that defining problem-specific grammars annotations to shared identifiers complying with the MIRIAM implementing custom hierarchies and part categorization (Minimal Information Requested In the Annotation of biochemical schemes facilitate the use of repository content for various Models) standard. applications. Examples of the implementation of this approach in Conclusions: Together with the offered web service access, the GenoCAD will be presented. consistent structuring and annotation of reaction kinetics data makes SABIO-RK well suited for its integration into workflow DS1-2-04 applications using or requiring kinetic data, such as modelling platforms in systems biology. Reactome - a knowledgebase of human biological pathways DS1-2-06 Schmidt, Esther1; Wu, Guanming2; Vastrik, Imre1; Croft, David1; de Bono, Bernard1; Gopinath, Gopal2; Gillespie, Marc2; Jassal, Bijay1; Reporting standards for omics data - synergistic efforts Matthews, Lisa2; Garapati, Phani1; Caudy, Michael2; Kanapin, Sansone, Susanna-Assunta Alexander2; Birney, Ewan1; D’Eustachio, Peter3; Stein, Lincoln2 The European Bioinformatics Institute (EBI - EMBL), Cambridge, 1EMBL-EBI, Cambridge, United Kingdom; 2Cold Spring Harbor United Kingdom Laboratory, Cold Spring Harbor, United States; 3NYU School of Medicine, New York, United States Objective: Fields of study such as transcriptomics, metabolomics, and proteomics are the primary source of data Reactome is a manually curated knowledgebase for human for the construction and validation of system biology models. It biological processes and currently contains ~ 2500 proteins is pivotal that such datasets are reported in a standard manner in 2400 reactions. Curation is based on direct interactions to enable communication, interpretation and analysis. To this with experts across numerous fields, resulting in a high quality end, standards for data content (minimal information checklists), network of molecular entities and their molecular transformations semantics (ontologies) and syntax (file formats) are being defined that can serve as a framework for Systems Biology analyses. within individual omics. However, reporting standards should Topics included in Reactome span metabolism, signalling stand alone but should also function well together to overcome pathways, transport processes, cell cycle, apoptosis, and more. fragmentation. In addition to the human events, the Reactome website features Results: Several synergistic activities have begun fostering the electronically inferred events to 22 non-human species, as well harmonization and consolidation of the three kinds of reporting as multiple cross-references to entries in other publicly available standards being developed. Over 20 projects are registered in data resources. The Reactome Mart provides access to powerful the “Minimum Information about a Biomedical or Biological” queries of Reactome data as well as combined explorations (MIBBI) portal [1,2] set to created orthogonal checklist modules. across other “martified” databases. The SkyPainter can be At present, over 60 groups participate under the OBO Foundry used to overlay user-supplied data on the Reactome reaction umbrella [3,4] with the objective of developing interoperable network, or to perform an overrepresentation analysis for a list ontologies. Several groups participate in the Functional Genomics of identifiers. All data is freely available and can be downloaded (FuGE) project [5,6] which underpins the XML-based formats in a number of common export formats including SBML. An they have developed. Only very recently, another complementary update on recent additions and ongoing projects will be provided, initiative has sprung up from a growing number of communities including descriptions of the much improved search results page, that work collaboratively on a common tabular framework for a functional interactions extension of the SkyPainter, obtained presenting the experimental metadata (ISA-TAB) [7,8].

ICSB 2008 17 Conclusions: Undoubtedly, interoperable standards will benefit escalated hugely, matched by a significant decline in the number the entire scientific community. They will also ease the task of of new medicines reaching the market. Further, compared with software developers, particularly those working to implement the 1960s, the time taken to develop a new drug has doubled standards-compliant systems for complex multi-omic studies, to approximately 12 years. The available data from across the such as the BioInvestigation Index at EBI [9]. Ultimately, industry shows an increased productivity in the discovery phase, standards-compliant data systems will be a valuable resource for but this has not been matched by success in development, with the system biology community. a number of well-publicised failures recently of drugs in the later 1. http://mibbi.sf.net stages of the development pipeline. This suggests that whilst 2. Taylor, Field, Sansone et al. Nat Biotechnol (in press). the output of projects from discovery into development has 3. http://www.obofoundry.org increased, the quality of the output has not. 4. Smith, Ashburner, Rosse, C et al. (2007). Nat Biotechnol. Pharmaceutical R&D generally has been an empirically data- 25(11):1251-5. driven, qualitatively oriented, activity. Whilst targets being studied 5. http://fuge.sf.org in Discovery may be placed within a network, it is not necessarily 6. Jones, Miller, Aebersold et al. (2007). Nat Biotechnol. clear which of the many targets is the one worth pursuing 25(10):1127-33. therapeutically, because of the complexity of the biological system Dedicated 7. http://isa-tab.sf.net itself, compounded by variability between individuals. Typically,

Orals 8. Sansone, Rocca-Serra, Brandizi et al. OMICS (in press). each drug and target combination tends to be considered in 9. www.ebi.ac.uk/net-project isolation, using target-driven, high throughput, approaches that are removed from their physiological context. The key weakness Dedicated session 1-3: Drug discovery of reductionism is that it cannot be used to understand or predict how the wider system will behave in a quantitative way. In other words the dynamic “context” (pathway, cell, patient or population DS1-3-01 of patients) is missing. Systems Biology is seen increasingly as an approach that can help tackle this problem, especially Targeting the networks and their fragilities in articles in the technical and industry press. It has also been Snoep, Jacky L.1; Bruggeman, Frank2; Bakker, Barbara2; the subject of recent UK and European initiatives. This talk will Westerhoff, Hans V3 offer a perspective on this topic, touching on opportunities and 1University of Stellenbosch, Biochemistry, Stellenbosch, South challenges that it presents, including some thoughts on industry/ Africa; 2Netherlands Institute for Systems Biology, Amsterdam, academia collaboration. Netherlands; 3Manchester Centre for Integrative Systems Biology and Netherlands Institute for Systems Biology, Astra Zeneca Chair DS1-3-03 for Systems Biology, Manchester, United Kingdom Parameterization of a large-scale, autonomous network Objective: Most diseases that are important today are model of the hepatic metabolism from transient metabolite multifactorial, i.e. network (systems) biology diseases. We aim data to digress from the existing strategies for drug (target) design Maier, Klaus1; Mauch, Klaus2; Hofmann, Ute3; Niebel, Anja1; and to devise new methodologies that truly target the diseased Vacun, Gabriele1; Reuss, Matthias1 networks. 1University of Stuttgart, Institute of Biochemical Engineering, Results: Using combinations of experimentation, modelling, Stuttgart, Germany; 2Insilico biotechnology, Stuttgart, Germany; modelbase exploration and theory, we have determined the 3Dr. Margarete Fischer-Bosch Institute of Clinical Pharmacology, robustnesses and fragilities in various intracellular networks. We Stuttgart, Germany found that networking increases robustness. Differential network based fragility analysis may identify drug targets that are effective Objective: Predictive dynamic network models of the liver against disease without leading to toxicity, at least in the tissues metabolism are of particular interest for quantitative systems- that can be analyzed presently. oriented analyses of detoxification processes and metabolic Tumorigenesis engaging the EGF receptor network appears disorders like, for example, diabetes. Furthermore, such models to increase robustness, making the amplified oncoprotein an can support prospective personalized prognoses of drug actions unattractive drug target. A conservation principle that we illustrate and/or their persistency, which accrues a reduced risk of severe numerically and prove analytically, removes the associated side effects. In this contribution, a first autonomous in silico pessimism vis-a-vis drug target identification: somewhere in network model identified from hepatic in vivo time-series data is the network, the tumor cells should be more fragile than their presented. untransformed counterparts. We show how to identify that better Results: Hepatocytes were incubated in standard six-well tissue drug target through Systems Biology. culture plates at 37°C in 5% CO2 atmosphere. Subsequently to The first phase of our analysis is for single functions of the perturbing the cells in a stimulus response experiment, extra- and network such as transcription activation through ERK-PP. We intracellular samples were collected in triplicate and subjected to shall also present the second phase, which deals with the absolute quantification by HPLC-, GC-MS- and LC-MS- analyses. simultaneous attack of two functions of the same network by a The observed transient metabolite data was then applied for single drug. And then a third phase deals with dual attack of one identifying metabolic network dynamics in the hepatic central or more functions. carbon metabolism. Canonical linear-logarithmic (linlog) kinetics Conclusions: New drug targeting approaches that target the were employed to approximate enzymatic reaction rates. Thus, network rather than a molecule have been developed. all rate equations share the same mathematical structure in which They lead to different predictions of optimal drug targets. metabolite and effector levels are considered by summation They take drug toxicity / drug safety into account. of logarithmic concentration terms. The kinetic parameters (i.e. elasticities) were estimated by matching the mathematical DS1-3-02 model to the experimentally observed metabolite data, that is by minimizing the variance-weighted sum of squared residuals Challenges and opportunities for systems biology in drug between measured and simulated metabolite time-series data. discovery: A personal perspective The parameter identification was realised by applying an evolution Henney, Adriano strategy together with a self-adaptive mutation operator. For Astra Zeneca, Manchester, United Kingdom efficient simulation the linearly-implicit differential algebraic solver LIMEX was invoked. Parameter estimation within the linlog The opportunities for Systems Biology in helping to bring framework allowed for straightforward computation of control innovative and effective new medicines to the market largely arise coefficients. from the challenges that the pharmaceutical industry now faces. Conclusions: The experimental data observed in the Over the last ten years the cost of developing new drugs has perturbation experiment offered adequate changes in metabolite

18 ICSB 2008 levels for identifying network dynamics. The transient intra- and DS1-3-06 extracellular metabolite time-series were successfully employed for a data-driven reconstruction and parameterization of a Compartmental modelling of the breast cancer resistance dynamic model of the hepatic metabolism. protein (BCRP/ABCG2) in drug transport Atari, Mohammed1; Chappell, Michael1; Errington, Rachel2; Smith, DS1-3-04 Paul2; Evans, Neil1 1University of Warwick, School of Engineering, Coventry, United Using physiology-based pharmacokinetic modelling to Kingdom; 2Cardiff University, School of Medicine, Cardiff, United dissect mechanisms in hepato-biliary transport Kingdom Kuepfer, Lars1; Goerlitz, Linus2; Lippert, Joerg1; Niederalt, Christoph1; Siegmund, Hans-Ulrich1 Objective: The human breast cancer resistance protein (BCRP/ 1Bayer Technology Services GmbH, Systems Biology, Leverkusen, ABCG2) of an ATP-binding cassette half-transporter is resistant Germany; 2Bayer Technology Services GmbH, Computational to topotecan (TPT) a semi-synthetic derivative of camptothecin. Solutions, Leverkusen, Germany Over-expression of BCRP is linked with high levels of resistance to the anti-cancer agent TPT by promoting drug efflux. Hepato-biliary transport is one of the pivotal pharmacokinetic To investigate this efflux pump mechanism, a compartmental steps which widely determines drug availability within the human model for the in vitro uptake kinetics of TPT has been extended body. Since the underlying processes have direct influence from a previously published model. The new model describes

on the efficacy and the safety profile of a drug they are of key the drug activity and delivery of the pharmacologically active Orals importance when developing new substances. It is hence crucial form to the DNA target as well as the elimination of drug from the Dedicated to quantitatively assess hepato-biliary transport mechanism in cytoplasm via the active pump. order to evaluate and analyse the elimination of new compounds. Results: Validation of the proposed model is achieved using Many of the main factors that influence transport within the scanning-laser microscopy data from live human breast cancer liver, however, are still ambiguous to date or can only be cells (MCF-7 cell line). The experiment used to collect these data quantified with great difficulty. We show here that physiology- imposes an output structure on the model that corresponds to based pharmacokinetic (PBPK) models together with targeted the functions of the variables that can be directly observed (total experimental data can unravel key operating mechanisms in concentrations of TPT in extracellular medium, cytoplasm and hepato-biliary drug transport. In particular, mechanistic whole- nucleus). Before estimating the unknown model parameters body PBPK models in combination with plasma distribution from the collected data it is essential to determine parameter data and expression profiles may be used to quantify the relative uniqueness (or otherwise) from this imposed output structure. impact of genetic and physiological variability on inter-individual This determination is formally performed as a structural changes in drug availability. Moreover, PBPK-modelling allows identifiability analysis, which demonstrates that all of the unknown to analyse the influence of polymorphisms in single transporter model parameters are uniquely determined by the output genes hence providing an important step towards individualized structure corresponding to the real experiment. medicine. The tools and approaches introduced will be Conclusions: The fitted model has shown that, as the exemplified for (co-) administration of various drugs, in particular concentration of the free transporter increases, the concentration Simvastatin and Gemfibrozil. of the active form of TPT bound to the transporter increases in the cell membrane, additionally, it reaches saturation at a higher DS1-3-05 rate than other compartments. However, the concentration of the inactive form of TPT bound to the transporter is at a lower A systems biology approach to cancer chronotherapy concentration in the cell membrane compared to the active form. Orrell, David; Fernandez, Eric; Hardy, Adam; Ramsell, Laura; Fell, Moreover, the model allows in silico estimations and predictions David; Chassagnole, Christophe of the relationship between the target binding and the dose, with Physiomics plc, Oxford, United Kingdom different expressions of the drug resistance protein, leading to the possibility of the design of optimal dosing regimens. Objective: Cancer chronotherapy is a treatment method in which the timing of drug delivery is set relative to periodic biological DS1-3-07 rhythms, such as the cell cycle, to give the maximum health benefit to the patient. As part of the TEMPO consortium, within Physiologically based pharmacokinetic modelling in the the EU’s Sixth Framework Programme, Physiomics has applied context of drug discovery: Melding pharmacokinetics with its mathematical models of the cell cycle to optimise delivery host-virus dynamics schedules of anticancer agents such as Seliciclib. The project von Kleist, Max; Huisinga, Wilhelm is currently limited to building in silico models of the mouse; Hamilton Institute, Dublin, Ireland however the ultimate aim of chronotherapy is to provide tailored drug schedules for individual patients. Objective: In HIV disease, the mechanisms of drug-resistance Results: Our pharmacodynamic (PD) cell cycle model, which are only poorly understood. The existence of latent viral reservoirs, uses circadian clock genes as its internal biological oscillator, in which drug treatment is sub-optimal, is suspected to be crucial. was coupled with a pharmacokinetic (PK) profile of the anticancer As a first steps towards elucidating the impact of drug kinetics on agents to provide a complete PK-PD analysis framework. virus evolution, we aim at analyzing the pharmacokinetics of a key Single-cell simulations showed that the effect of a single dose class of anti-HIV drugs, so-called nucleoside reverse transcriptase of Seliciclib was highly dependent on the stage of the cell inhibitors (NRTI), in particular zidovudine (AZT). Currently, there is cycle. Multi-cell simulations of synchronous and asynchronous no model to analyse and predict the non-linear and time-delayed populations, which reflect the difference between healthy and relationship between the concentrations of the parent compound tumour tissues, were also performed using the University of in blood plasma and the active compound at the side of action. Swansea’s IBM “Blue C” supercomputer. These showed that Results: In the case of the NRTI zidovudine (AZT) di- specific administration times were crucial to obtain the maximum phosphorylation has been identified as the rate-limiting step in effect (i.e. cell cycle arrest) in tumour tissue, and the minimum CD4-cells (PBMC). We have established a mechanistic model effect in healthy, synchronised tissue. The results were in that describes the pharmacokinetics and the effect of AZT agreement with experiments conducted by TEMPO consortium at different doses. We found that AZT exerts its intracellular members. effect directly through its triphosphate and indirectly thorough Conclusions: Physiomics’ technology allows the simulation of its monophosphate. Model reduction allows for relating the interactions between the circadian clock, the cell division cycle, intracellular effect of AZT in CD4-cells to its blood plasma levels. and PK processes. The results showed that a systems biology We found that the intracellular activation of AZT is saturable in approach can be used to determine optimal drug schedules, and CD4-cells, so that by increasing the doses beyond a certain paves the way for the chronotherapeutic delivery of cancer drugs. threshold disproportional (and unfavorable) efficacy-to-toxicity

ICSB 2008 19 situations arise. an established colorimetric assay and an on-line respiration Conclusions: HIV exhibits a broad cell tropism so that due to the measurement in 96-well microtiter plates [1]. In addition, we mode of activation and transport of AZT in cells the presence of characterized metabolic fluxes using metabolite balancing and cellular latent reservoirs is very likely for this drug. By relating the 13C labeling. With Tacrine we identified clear differences in some intracellular effect of AZT in CD4 cells to its blood plasma levels, fluxes, e.g. pyruvate to alanine, in a subtoxic range, compared the model can be linked to models of within host dynamics of with other drugs such as Amiodarone that did not cause any HIV in order to elucidate the role of pharmacokinetics at different significant change. dosing schemes and the role of potential latent reservoirs. Conclusions: In our study we could show that metabolic flux analysis in Hep G2 cells permits the identification of effects of DS1-3-08 drugs that are clinically known to be toxic but do not show toxicity in conventional in-vitro pre-clinical screening methods in the A domino effect in drug action: From metabolic assault via subtoxic range. gene expression to parasite differentiation References: Haanstra, Jurgen1; van Tuijl, Arjen1; Blits, Marjolein1; Kerkhoven, [1] Deshpande RR, Kirsch Y, Maas R, John GT, Krause C, Heinzle Eduard1; van Nuland, Rick1; Albert, Marie-Astrid2; Michels, Paul2; E (2005) Microplates with Integrated Oxygen Sensors for Kinetic Dedicated Bouwman, Jildau1; Westerhoff, Hans1; Bakker, Barbara1 Cell Respiration Measurement and Cytotoxicity Testing in Primary

Orals 1Vrije Universiteit Amsterdam, Molecular Cell Physiology, and Secondary Cell Lines. ASSAY Drug Devel. Technol. 3: 299- Amsterdam, Netherlands; 2De Duve Institute, Research Unit for 307. Tropical Diseases, Brussels, Belgium Dedicated session 1-4: Plant systems Cells adapt to external perturbations, such as stress or drugs, by adaptations of metabolism and gene expression. Regulation Analysis has shown that regulation is distributed between DS1-4-01 the gene-expression cascade and metabolism and how this distribution can be quantified [1]. Even within the gene-expression Systems biology in Arabidopsis and Chlamydomonas. cascade regulation is distributed with regulation taking place Integrating transcription, transaltion, metabolism and not only at transcription, but also at mRNA processing and growth stability and protein synthesis, modification and breakdown [2, Stitt, Mark 3]. These new insights should have their impact on systems- Max Planck Institute of Molecular Plant Physiology, Potsdam- based identification of drug targets. Awareness is growing Golm, Germany that drug-target validation should involve system analysis of cellular networks. There is less appreciation, however, that Objective: Plants grow continuously changing conditions. expression of these networks may change in response to drugs. Every day they alternate between photosynthesis in the light and Glycolysis is the sole source of free energy for the deadly parasite respiration in the dark. Conditions also change from day to day, Trypanosoma brucei. Plasmamembrane glucose transport and on a seasonal basis. We want to understand how plants exerts a high control over trypanosome glycolysis and hence is gauge their rate of growth to fluctuating resources. Our starting a promising drug target. We show that inhibition of trypanosome point is to ask how they balance theircarbon budget over a 24 glucose transport initiates a domino effect in which network hour cycle. Some photosynthate is stored as starch in the light, adaptations enhance the primary inhibition via (i) metabolic and remobilised at night to support respiration and growth. This control, (ii) downregulation of the expression of the primary target is precisely regulated, so that starch just lasts till dawn. We have and its pathway and (iii) expression of an insect-stage coat protein accumulated a large body of data about transcript levels, enzyme that should be immunogenic in humans. This may open a new activities, polysome loading, metabolite levels and growth rates, avenue in systems-biology based drug target discovery, aiming at in large numbers of different genotypes. I will present models that tricking parasites into lethal adaptation strategies. integrate this data in different ways. References: 1. Rossell, S., et al. (2006) Proc. Natl. Acad. Sci. Results: The first is a simple linear model that predicts global 103, 2166-2171 changes in gene expression during diurnal cycles, using three 2. Daran-Lapujade, P. et al. (2007) Proc. Natl. Acad. Sci. 104, inputs; the clock, light and sugars. This model shows that sugars 15753-15758 interact closely with the other two inputs and play a major role in 3. Haanstra, J. et al. (2008) J. Biol. Chem. 283, 2495-507 the diurnal regulation of gene expression in Arabidopsis leaves. The second address the impact of these changes of transcripts DS1-3-09 on enzyme levels in central metabolism. Quantitative analyses of polysome loading allow us to estimate the rates of synthesis Metabolic flux analysis in Hep G2 cells indicates effects of of enzyme proteins. These are compared with measurements drugs in subtoxic range of protein levels to identify enzymes that are subject to rapidly Heinzle, Elmar; Noor, Fozia; Niklas, Jens turnover, and enzymes where turnover is much slower and Saarland University, Biochemical Engineering, Saarbrucken, transcriptional regulation serves to allow slow adjustment to Germany prolonged changes in the environment. The third is a model of whole plant growth, which explores the consequences of altering Objective: Early prediction of efficacy and adverse effects of starch accumulation and protein turnover for the rate of plant new chemical entities is of major importance in drug discovery growth. and development. Since 80% of pharmaceutical costs are in Conclusions: Models provide a vital tool to integrate, and the clinical development, good filtering out of drug candidates explore complex functional data sets, and draw biological in preclinical development is essential to avoid late attrition messages from them. and failure. Existing methods are limited in their predictability, emphasizing the need for new and better methods. Systems DS1-4-02 biology offers a new approach towards improving targets and characterizing toxicity by identifying biomarker proteins and From molecules to whole plants: Towards an integrative metabolites that can report on the behavior of the system. Our model of Arabidopsis thaliana hypothesis is that effects below toxic levels contribute to the later Prusinkiewicz, Przemyslaw failure of drugs due to toxicity. University of Calgary, Department of Computer Science, Calgary, Results: We therefore, developed and applied methods of Alberta, Canada metabolic flux analysis to the human hepatoma cell line, Hep G2, exposed to clinically used drugs some of which were withdrawn Objective. A major research objective in plant biology is the from the market due to adverse effects such as Troglitazone. In understanding of plant development in molecular terms. The talk a first step we identified subtoxic range of these drugs applying will describe recent results pertinent to the development of the

20 ICSB 2008 branching architecture of Arabidopsis thaliana. Calvin cycle. These are a) the rate at which carbon is withdrawn Results. Data-driven simulation models of spiral phyllotaxis, for the synthesis of carbohydrates and other compounds via vascular pattern development and bud activation have been the triose-phosphate-translocator b) light, mediated by electron created. The models support the hypothesis that the plant transport and its interactions with events in the stroma, and last c) hormone auxin plays a significant regulatory role in plant the supply of. development. Auxin is produced in the vicinity of the shoot apical Although there already exist detailed kinetic models of the meristem and transported in the epidermis toward the peripheral cycle, we are, in cooperation with experimenters, developing a zone of the apex. There it accumulates in emergent convergence simpler model, which may be much easier to analyze. We are points, which determine the arrangement of the incipient leaves, mainly focused on mechanisms, which have regulatory effects flowers, and axillary buds around the plant axis (phyllotaxis). on the cycle, such as inhibition and activation mechanisms. From the primordia auxin flows into the subepidermal layers of We analyze the effects of these regulatory mechanisms on the the apex and, eventually, into the plant stem. In this process, it is kinetic behaviour of the cycle in a systematic way to understand canalized into narrow paths that define plant vasculature. Within how the efficiency of the cycle is optimized. We are in particular the stem, auxin regulates the activation of lateral buds, and thus interested in the kinetic behaviour of the carbon translocator and coordinates the development of branches. the electron transport driven by light. One of the problems of the Conclusions. The key processes that underlie the development model is to bring the reactions of the electron transport, which of branching plant architecture are likely unified by the regulatory are very fast, in line with the reactions of the cycle, which are role of auxin. Computational models elucidate a plausible causal comparably slow.

structure of these processes and relate molecular mechanisms Orals operating at the level of plant cells to macroscopic plant DS1-4-05 Dedicated structures. The self-regulated polar transport of auxin represents a novel mechanism of morphogenesis, distinct from the established Integration of light temperature signalling paradigms of positional information and reaction-diffusion. Halliday, Karen; Foreman, Julia; Millar, Andrew; Sorokin, Anatoly; Goryanin, Igor DS1-4-03 Edinburgh University, Biological Sciences, Edinburgh, United Kingdom Theoretical modelling reveals competitive complex formation as the core of an activation-depletion framework Objective: In the natural environment plants are subject to Geier, Florian1; Digiuni, Simona2; Schellmann, Swen2; Greese, changes in ambient light levels, light quality and daily fluctuations Bettina3; Pesch, Martina 2; Wester, Katja2; Dartan, Burcu2; Mach, in temperature. We want to understand how these environmental Valerie2; Sriniva, Bhylahalli Purushott2; Timmer, Jens3; Fleck, inputs are integrated at the molecular level. Christian3; Hulskamp, Martin2 Results: Our work has shown that the molecular response 1University of Freiburg, Institute of Physics, Freiburg, Germany; to these two environmental signals is linked. For instance, the 2University of Cologne, Cologne, Germany; 3University of Freiburg, collective action of the phytochrome light receptors buffers effect Freiburg, Germany of temperature on flowering time and growth (Halliday at el, 2003; Halliday and Whitelam, 2003; unpublished data). Our work has Objective: Trichome patterning in Arabidopsis serves as a model established that a specific transcription factors (TFs) in the basic system for de novo pattern formation in plants. It is thought to Helix-Loop-Helix (bHLH) class operate in a coherent feed forward typify the theoretical activator-inhibitor mechanism although Module (CFFM) to integrate light and temperature information this hypothesis has never been challenged by a combined (Penfield et al., 2005; unpublished data). We have characterised experimental and theoretical approach. a CFFM that provides the cool arm of a growth buffering Results: By integrating the key genetic and molecular data of mechanism (that operates between 17C-27C). the trichome patterning system we develop a new theoretical Conclusions: The cool CFFM, suppresses cell expansion in the model which exhibits properties that cannot be reproduced by 17C-22 C range. We hypothesise that molecular substitution any previous model. We show experimentally that the trichome of bHLH TFs, will alter the functional temperature range of the inhibitor TRIPTYCHON is transcriptionally activated by the CFFM, providing a mechanism for temperature buffering. The known positive regulators GLABRA1 and GLABRA3. Further, by involvement of molecular isoform switching in circadian clock particle bombardment of protein fusions with GFP we show that temperature compensation provides the possibility that this TRIPTYCHON and CAPRICE but not GLABRA1 and GLABRA3 may be a common buffering mechanism. In our newly (BBSRC- can move between cells. Finally, theoretical predictions suggest EPSRC) funded ROBuST project we will combine experimental promoter swapping and basal over-expression experiments by and theoretical approaches to assess the impact of temperature means of which we are able to discriminate three biologically on a defined network. This will provide a more comprehensive meaningful variants of the trichome patterning model. view of network robustness. Conculsion: Our work demonstrates that the mutual interplay between theory and experiment can reveal a new level of DS1-4-06 understanding of how biochemical mechanisms can drive biological patterning processes. Growth patterns - modeling of morphogenesis in shoot apical meristem DS1-4-04 Krupinski, Pawel1; Jonsson, Henrik1; Couder, Yves2; Boudaoud, Arezki2; Hamant, Olivier3; Heisler, Marcus4; Sahlin, Patrik1; Traas, Kinetic modeling of the Calvin cycle Jan3; Meyerowitz, Elliot4 Solmaz, Erim; Ebenhöh, Oliver 1Lund University, Computational Biology and Biological Physics, Max Planck Institute of Molecular Plant Pysiology, Potsdam, Lund, Sweden; 2l’Ecole Normale Supérieure, Département de Germany Physique, Paris, France; 3UMR INRA-CNRS-ENSL-UCB Lyon I, Laboratoire de Reproduction et Développement des P, Lyon, The regulation of the pentose phosphate pathway (the Calvin France; 4California Institute of Technology, Division of Biology, cycle) for photosynthetic carbohydrate formation inplants has Pasadena, United States been the subject of extensive experimental and theoretical studies. The reactions of the cycle occur in the chloroplast Objective:The Shoot Apical Meristem (SAM) is a population stroma. The cycle itself comprises 13 reactions, catalyzed by of mitotic cells initiating new aerial organs of a plant. Extensive eleven enzymes and has three phases. It is dependent on input genetic analysis has unraveled a complex network of regulatory processes supplying ATP, NADPH and to the reaction system, as factors controlling meristem function, however it remains unclear well as on output processes withdrawing cycle intermediates from how this network translates into SAM and organ morphogenesis. the system. In the past, several theoretical frameworks have been investigated There are three factors determining the overall turnover rate of the as to how mechanical forces might induce differential growth

ICSB 2008 21 patterns and thus contribute to morphogenesis. Here we have phosphenolpyruvate translocators are also involved in the transfer used modeling approach to explore this question further. of redox equivalents across the chloroplast envelope. Results: From a mechanical point of view, the shape of a tissue is determined by equilibrium of forces resulting from turgor, DS1-4-08 stresses in the cell walls and symplastic growth. This picture is complicated by the fact that cell walls achieve a large degree of Stomatal development in Arabidopsis leaf development anisotropy due to the presence of an oriented cellulose fibrillar Kheibarshekan ASL, leila1; Govaerts, Willy1; Dhondt, Stijn2; network, and that the mechanical properties of the cell wall are Boudolf, Véronique2; De Veylder, Lieven2 changing during the course of plant development, notably via the 1Gent University, Applied Mathematics and Computer Science, microtubular network which controls the oriented deposition of Ghent, Belgium; 2Ghent University, VIB,Department of Plant the cellulose microfibrils. Systems Biology, Ghent, Belgium To analyze the role of mechanical stresses, strains and molecular patterns in morphogenesis, we compare the results of simulation Objective: The abaxial epidermis of the first leaves of Arabidopsis with live imaging data obtained by confocal microscopy. In consists of 2 cell types, stomatal guard cells and pavement cells. addition to parameters related to cell physiology, the mechanical Stomata are small pores on the surface of leaves whose aperture Dedicated part of the model includes walls as structurally strong elements is controlled by 2 guard cells. When these guard cells are open,

Orals in two or three dimensions using spring network models or the stomata allow gas exchange, mainly CO2 for photosynthesis finite element models. At each step of the simulation, changes and H2O, between the leaf and the atmosphere. Here we in mechanical conformation of the cells result in changes of concentrate on the initiation and regulation of precursor cells that strain and stress patterns which in turn influence the internal cell form guard cells and pavement cells, leading us to approximately properties like microtubule-driven anisotropy, turgor pressure or compute the probabilities of the pathways. This paper will restrict morphogens concentrations. primarily to the number of pavement and guard cells. However Conclusions: We will show how our mechanical models of the some of the other cells which exist in the pathways will be used SAM are able to recapture patterns of microtubules alignment for the interpretation of the stomatal pathway. experimentally observed by confocal microscopy during normal Results: In this paper besides the probabilities of the pathways shoot and organ outgrowth, and in a context where stress we also compute the cell cycle duration and we study the relation and strain have been altered following cell ablation, providing between cell area and the onset of the endocycle. The idea is plausible mechanisms for the control of microtubules orientation that when a cell enters the endocycle, then it stops dividing and and its impact on cell wall anisotropy, primordia formation and keeps growing. We estimate the threshold for the cell area above morphogenesis. which cells do not divide anymore. Furthermore, we compute the growth rates by using the threshold during leaf development. DS1-4-07 Conclusions: It is possible to compute the probabilities for dividing into pavement cells or forming stomata guard cells from Construction and analysis of a model of plant carbon experimental data. A most interesting result of this work is that the metabolism probability producing stoma is a nearly fix small number. Chokakthukalam, Achuthanunni1; Poolman, Mark2; Sweetlove, Lee3; Fell, David2 Dedicated session 2-1: 1Oxford Brookes University, Cell Systems Modelling Group, Oxford, United Kingdom; 2Oxford Brookes University, Cell Cell-regulation - signalling Systems Modelling Group, Oxford, United Kingdom; 3Oxford University, Plant Sciences, Oxford, United Kingdom DS2-1-01

Objective: The metabolism of a plant cell is distributed between Linking dynamic properties of signaling networks with various compartments e.g. the cytosol, chloroplast and the cellular decisions mitochondria. Each compartment has specific functions and Klingmüller, Ursula hence has its own set of reactions. Transporters mediate German Cancer Research Center, Heidelberg, Germany the transfer of metabolites between compartments. It has been shown that two metabolite shuttles can operate across Cell growth and differentiation processes are tightly controlled the chloroplast membrane: the malate-oxaloacetate shuttle by the activation of complex intracellular signaling networks. We (MAL/OAA shuttle) and the dihydroxyacetone phosphate-3- combine quantitative data generation and dynamic pathway phosphoglycerate shuttle (DHAP/PGA shuttle) to transfer redox modeling to identify general design principles of submodules potential from the chloroplastic NADPH to the cytosolic NAD. of signaling networks in erythroid progenitor cells and primary The presence of other transporters on the chloroplast membrane hepatocytes. By standardizing cell systems and experimental raises the possibility of additional potential routes of transfer of procedures we could establish data-based mathematical models redox potential across chloroplast envelope. for JAK2-STAT5, Smad, PI3 kinase and MAP kinase signaling. Models describing biochemical reaction systems can be We could show that the stoichiometry of pathway components constructed based on their stoichiometric coefficients, known as and the arrangement of negative feedback loops critically shape stoichiometric or structural models. Various model interrogation the dynamic properties of signaling network. Time-resolved techniques like elementary modes (EMs) analysis can be used expression profiling in combination with physiological read outs to reveal properties of such networks. An EM is a minimal set enabled us to link our dynamic data-based mathematical models of enzymes that can operate at steady state with all irreversible with cellular responses. The quantitative prediction of responses reactions working in the thermodynamically favoured direction. in healthy mammalian cells will provide an important basis to EMs analysis can be used to identify productive routes within a elucidate alterations promoting the development of cancer. network. The objective of this study is to identify the presence and significance of such routes using EMs analysis. Results: A model of plant metabolism has been constructed by extending a previous model of Calvin cycle by including the light reactions and cytosolic glycolylis. Interrogation of the model using EMs analysis has shown a number of routes by which redox potential may be transferred across the chloroplast membrane without any net carbon flux. These modes involve oxidation of NADPH in the stroma accompanied by corresponding reduction of NAD in the cytosol. Conclusion: It was observed that in addition to MAL/ OAA and DHAP/PGA shuttles, glucose 6-phosphate and

22 ICSB 2008 DS2-1-02 Results: The model is fitted to data for different conditions as well as different mutants, providing evidence that the most important Spatial and temporal information encoding by the NF- processes are well captured by the model. The fitted model is kappaB system used to explore the role of the two branches by simulations, White, Michael1; Horton, Caroline2; Nelson, David2; Paszek, making predictions for further experiments. The model shows Pawel2; Ashall, Louise2; Sillitoe, Kate2; See, Violaine2; Spiller, that the Sln1 branch is more sensitive then the Sho1 branch, David2; Harper, Claire2; Ryan, Sheila2; Kell, Douglas3; Broomhead, the latter only reacting above an osmotic shock of 0.1 M NaCl. David4; Rand, David5 However, the responses of both branches are quickly saturated. 1University of Liverpool, School of Biological Sciences, Liverpool, Interestingly, the Sho1 branch activates the MAPK Hog1 to a United Kingdom; 2University of Liverpool, School of Biological lesser extent than the Sln1 branch at low stress, even though the Sciences, Liverpool, United Kingdom; 3University of Manchester, adaptation time is similar. Thus, at low stress Hog1 activation and School of Chemistry, MIB, Manchester, United Kingdom; subsequent increased enzyme translation and glycerol production 4University of Manchester, School of Mathematics, Manchester, is not the main mechanism of osmo-adaptation but rather glycerol United Kingdom; 5University of Warwick, Systems Biology Centre, accumulation by glycerol-channel closure. Coventry, United Kingdom Conclusions: This modelling study gives new insights into yeast osmo-adaptation. Yeast maintains a rather high steady state Objective:Oscillations in NF-kappaB translocation are driven glycerol production to be able to quickly respond to low osmotic through a negative feedback loop through IkappaBalpha changes in the environment just by channel closure. The Hog1-

transcription. We have used single cell imaging to investigate mediated gene expression response plays no role in short term Orals this system in single cells. Timed stimulation of cells with pulses osmo-adaptation, but might prepare the cell for subsequent Dedicated of Tumour Necrosis Factor alpha (TNFalpha) were used to stress by up-regulating the steady-state glycerol production. At investigate the response of the NF-kappaB system. The aim higher osmotic perturbations, however, both channel closure was to test existing models with these data and to study the and increased glycerol production contribute to the adaptation downstream biochemical and gene transcription consequences of response. these repetitive pulses and their frequency. Results: We found that the system has a reset time after which DS2-1-04 a full amplitude pulse of NF-kappaB nuclear occupation is observed when a 5 min pulse of TNFalpha is applied. Repetitive Shared transcription programs for miRNAs and their 60, 100 or 200 min pulses led to synchronous oscillations in the targets is a common feature in the mammalian regulatory cell population assisting biochemical analysis (Western blotting, network ChIP or qPCR analysis of target gene expression). These studies Shalgi, Reut1; Oren, Moshe2; Pilpel, Yitzhak1 confirmed cyclical phosphorylation at RelA Ser536. Refitting 1Weizmann Institute of Science, Molecular Genetics, Rehovot, computational models to these data predicted the role of the A20 Israel; 2Weizmann Institute of Science, Molecular Cell Biology, feedback loop in NF-kappaB signalling. Varying pulse frequencies Rehovot, Israel had differing effects on early and late NF-kappaB-dependent gene expression. We also used a stochastic 3 feedback Objective: Can cells plan ahead? Can a cell program, upon computational model to suggest that noise in IkappaBepsilon transcription activation of genes, when and how they will be expression may have a role in increasing cell to cell oscillation shut off? These questions lead us to examine the possibility asynchrony. of coordination between transcription and post-transcription Conclusion:These studies suggest a role for stochastic regulation. More specifically we ask if miRNAs (miRs), prominent transcription in controlling cell to cell variation and showed post-transcriptional regulators, are involved in a cross-talk with that cells undergo repetitive cycles of RelA phosphorylation the transcriptional program in a mammalian cell. and dephosphorylation. We tested and disproved previous Results: Analyzing several mammalian genomes we find that computational models, making new testable predictions. Finally, in many cases miRs do not act alone. They are embedded in these data suggest that the frequency of repeated oscillations the network in regulatory circuits, which involve transcription may control differential NF-kappaB-dependent gene expression. regulation. Feed-forward loops (FFLs) in which miRs and their targets share a common transcription program are over- DS2-1-03 represented in the mammalian network of transcription and post-transcripion regulation. Common regulators, miRs and Modelling yeast osmo-adaptation: Integration of two transcription factors (TFs) in such FFLs show a tendency for co- signalling branches expression or expression avoidance across tissues, suggesting Schaber, Joerg1; Ammerer, Gustav2; Grotli, Morten3; Medrala, a variety of functional roles in the network, as temporal switches Darmara4; Morillas, Montse5; Nordlander, bodil4; Pelet, Serge6; for pathways on the one hand, or in the creation of spatial Petelenz, Ela4; Peter, Matthias6; Posas, Francesc5; Klipp, Edda7; boundaries, during development for example, on the other hand. Hohmann, Stefan4 This finding adds another level to the concept of spatio-temporal 1MPI Molecular Genetics, Berlin, Germany; 2University of Vienna, avoidance of miRs and their targets previously introduced by Department of Biochemistry, Vienna, Austria; 3Department of Stark et al. and Farh et al., hinting at the role of TFs as third- Biochemistry, Göteborg University, Sweden, Göteborg, Sweden; party players in this design. We now experimentally investigate 4Department of Cell and Molecular Biology/Microbiology, predictions raised in the project. Among these is a FFL involved Göteborg, Sweden; 5Universidad Pompeu Farba, Departament de in the regulation of cell cycle and proliferation. Thus we gain new Ciències Experimentals i de la Salut, Unitat de Senyalització Cel. insights on the biological roles that such regulatory architecture lular, Barcelona, Spain; 6ETH, Institute of Biochemistry, Zurich, can play in mammalian cells. Switzerland; 7Max-Planck Institute for Molecular Genetics, Berlin, Conclusions: We show that miRs and their targets often have Germany shared transcriptional programs, and that coordination between transcription and post-transcription regulation characterizes the Objective: The yeast Saccharomyces cerevisiae activates the mammalian regulatory network. Co-expression or expression high-osmolarity glycerol (HOG) pathway to adapt to hyperosmotic avoidance between regulators could yield different regulatory conditions1. The core element is the mitogen activated protein consequence. This architecture might be beneficial in cellular kinase (MAPK) Hog1 that regulates gene expression, protein regulatory networks. activity and cell cycle. Upstream of Hog1 there are two redundant signalling branches, i.e. the so-called Sln1 branch and the Sho1 branch, which converge at the common MAPK kinase Pbs2. We present a mathematical model of the HOG pathway including the two branches. The objective of this work to complete the picture of the dynamics of yeast osmo-adaptation.

ICSB 2008 23 DS2-1-05 The approach is applied to unravel the differences between primary and cancerous hepatocytes. A compendium of 6,560 Cellular fluctuation in protein concentration and its role on experimental points was generated in primary hepatocytes and adaptive response the hepatocellular carcinoma cell line HepG2. Both cell types were Yomo, Tetsuya1; Tsuru, Saburo1; Ichinose, Junya2; Kaneko, stimulated with 10 different cytokines, in orthogonal combination Kunihiko3; Ying, Bei-Wen1 with 7 small-molecule inhibitors targeting key signaling mediators. 1Osaka University, Osaka, Japan; 2ERATO, JST, Osaka, Japan; For these conditions, the phosphorylation levels of 17 readouts 3The University of Tokyo, Tokyo, Japan were measured. Based on this data and a priori knowledge retrieved from curated databases, we were able to delineate Objectives: Noises in living systems play an important role in models specific for both primary and cancer hepatocytes. By biological processes. So far, the intrinsic noise in gene expression comparing the models we could uncover significant differences has been extensively elucidated, whereas, the origin of the in the signaling networks, in particular involving IGF-mediated extrinsic noise that is dominant in the total fluctuation in the signaling. Furthermore, data that could not be reconciled with the cellular concentration of gene products is scarcely identified. a priori knowledge pointed at potential new connections involving Here, we first identify the governing source of the global noise crosstalk between inflammatory and survival pathways. Dedicated in cellular protein concentration, then study the resultant Conclusions: Boolean-based methods allow to compare

Orals biological function on adaptation, as the phenotypic diversity in network topology with high-throughput data unraveling novel the genetically identical cell populations may cause the adaptive mechanistic insights. response to environmental changes. Results: We introduced an artificial gene expression circuit into DS2-1-07 the E. coli genome and refined a statistic model to investigate the origin of the extrinsic noise. Intriguingly, the noise in cellular Predictive modulation of human blood stem cell fate by the growth rate was found to be the dominant contributor to the systematic manipulation of cell-cell interaction networks fluctuation in cellular protein concentration, particularly at the Kirouac, Dan1; Ito, Caryn2; Yu, Mei1; Zandstra, Peter1 high level of gene expression. As the growth rate of cell volume 1University of Toronto, Biomaterials and Biomedical Engineering, works as a governing factor of dilution, the noise in growth rate Toronto, Canada; 2Insception Biosciences, Toronto, Canada perturbs the concentration of the entire cellular components that determines the growth rate of the next generation. Subsequently, Objectives: Stem cells are defined by their dual capacity we integrated a series of gene expression cassettes of the same for self-renewal and multi-lineage differentiation, serving design applied as described into the E. coli genome, where the as a foundational tool for regenerative medicine and tissue corresponding gene under the regulation of its native operon is engineering. Advancing these fields will require an in-depth deleted, to examine the fluctuation in these gene products toward understanding of the mechanisms that control stem cell fate, and the environmental adaptation. Fascinatingly, the expression of the the development of novel tools to translate this understanding target genes increased in response to the changing environment, into effective technologies. We are using the human blood in the absence of their native regulatory mechanisms that involved forming (hematopoietic) system as an experimental paradigm. in the response to the corresponding environmental changes. Our objective is to understand the cell interaction networks that Conclusion: The negative feedback relationship between regulate blood stem cell growth, and use his understanding the growth rate and the concentration of the whole cellular to design a novel culture technology (bioprocess) to generate components allows the phenotypic fluctuation regardless increased numbers of transplantable human blood stem cells for of the perturbation in cellular growing process. As well, the cellular therapy applications. cellular fluctuation in protein concentration could play a role on Results: Our approach has been to analyze, model, and adaptive responses. Taken together, these studies may provide a manipulate the dynamic cell-cell signaling networks established primitive design principle of a robust growing system against the between blood stem cells and their differentiated progeny. Inter- fluctuating environment. and intra-cellular networks associated with supportive and non- supportive blood stem culture conditions were reconstructed DS2-1-06 by dynamic profiling of cell sub-population gene expression and protein secretion. Bioinformatic integration with multiple Interrogation of the topology of signaling networks with biological databases has yielded cell population interactions, large-scale data within a boolean framework uncovers signaling pathways, and novel secreted ligands associated mechanistic differences between primary and cancerous with stem cell self-renewal and differentiation. These findings hepatocytes are being integrated into a differential equation-based model Saez-Rodriguez, Julio1; Alexopoulos, Leonidas1; Regina, of hematopoiesis wherein self-renewal and differentiation are Samaga2; Klamt, Steffen2; Lauffenburger, Douglas3; Sorger, Peter1 regulated by feedback signaling between mature cells and 1Harvard Medical School/M.I.T., Boston, United States; 2Max progenitors. Planck Institute for Dynamics of Complex Technical Systems, Conclusions: Our integrated approach has identified novel Magdeburg, Germany, Magdeburg, Germany; 3M.I.T., Cambridge, regulatory axes wherein self-renewal is modulated by inhibitory United States chemokine-macrophage interactions, and stimulatory growth factor-megakaryocyte interactions. By modeling these interactions Objective: Novel high-throughput methods allow the generation we are able to accurately replicate blood stem cell growth and of large data sets, which are the ideal substrate for a system- differentiation under a variety of conditions, and perform in silico wide analysis. Sophisticated insight can be obtained with process design and optimization. Dynamic culture manipulations physicochemical models, but modeling large networks is unfeasible such as media dilution and cell population selection have been . Intermediate approaches based on Boolean (logical) descriptions incorporated into a clinical bioprocess capable of inducing can be useful, as they encapsulate the topology and causality of the expansion of transplantable human blood stem cells. network without dealing with kinetic parameters. Results: We use this framework to test the consistency between the literature-mined network topology and experimental data across different cell types, conditions, and time scales. Furthermore, we have developed a methodology to identify the topology that optimally describes a data set based on a priori knowledge of the network connectivity. These methods are embedded in CellNetOptimizer (CNO), an open source MATLAB toolbox that uses CellNetAnalyzer as simulation engine. CNO works in concert with DataRail (Saez-Rodriguez et al., Bioinf. 24:6 2008) a complementary toolbox for managing, transforming and visualizing data.

24 ICSB 2008 DS2-1-08 in addition a Bayesian approach to derive parameter values both from general knowledge about parameter value distributions and Computational analysis of effects of DNA damage from specific measurements2. response on circadian rhythms Conclusions: A number of approaches have been undertaken to Hong, Christian I1; Zamborszky, Judit2; Csikasz-Nagy, Attila2 reduce the uncertainty on various levels of biochemical network 1Dartmouth Medical School, Department of Genetics, Hanover, modeling as well as to incorporate the remaining uncertainty in United States; 2The Microsoft Research, University of Trento order to deduce the uncertainty in the respective predictions. CoSBi, Trento, Italy 1Liebermeister W & Klipp E, Bringing metabolic networks to life: convenience rate law and thermodynamic constraints. TBioMed, Objective: Cell cycle and circadian rhythms are conserved 2006. 3. biological processes with robust cyclic features. The interactions 2Liebermeister W & Klipp E, Bringing metabolic networks to life: between these two oscillators are objective of active research. integration of kinetic, metabolic, and proteomic data. TBioMed, Cell division cycle seems to be gated by circadian clock, possibly 2006. 3. by the clock transcription factor (BMAL1/CLK) induced cell cycle kinase, WEE1. On the other hand, a clock component (i.e. FRQ DS2-2-02 in Neurospora crassa and mPer1 in mouse) is phosphorylated by cell cycle kinase, Chk2, upon DNA damage. These data indicate Modelling with identifiable models that there are bi-directional interactions between the cell cycle Timmer, Jens; Hengl, Stefan; Kreutz, Clemens; Maiwald, Thomas

and the circadian clock. Experiments reveal that ionizing radiation University of Freiburg, Institute of Physics, Freiburg, Germany Orals treatments create predominantly phase advances in circadian Dedicated rhythms. Objective: Mathematical models of the dynamics of cellular Results: The molecular details of how DNA damage does not processes promise to yield new insights into the underlying cell cause relevant delays, remains uncovered. In order to address biology and their systems’ properties. Since the processes are this, we employ our previous simple mammalian circadian clock usually high-dimensional and time-resolved experimental data model to study possible Chk2 dependent molecular mechanism of the processes are sparse, parameter estimation faces the that results the observed behavior. We hypothesize different Chk2 challenges of structural and practical non-identifiability of the targets of PERs (monomer, dimer, and complex with BMAL1/ parameters. Non-identifiabilities might render the systems analysis CLK), and find that the unique phase advances of the circadian of the model difficult. Non-identifiability results in, in general, clock from DNA damage response are observed when Chk2 only non-linear dependencies of the estimated parameters. To infer affects PERs that are not bound to BMAL1/CLK. (non-)identifiability elegant analytical approaches exist which are, Conclusions: The implications of the advanced phase responses however, due to their computational complexity limited to low- of the circadian clock are to provide a mechanism to ensure dimensional systems. Established methods for high-dimensional more time for DNA repair. Our simulations suggest that Chk2 systems rely on linear approximations which renders the dependent phase advance is the best strategy to induce large interpretation of their results difficult. induction of WEE1 that inhibits entry into mitosis. Cell cycle Results: We show that identifiability analysis can be reduced network is ingeniously wired with circadian rhythm for an optimal to an intuitive geometric issue. To operationalise this intuition, result upon DNA damage. we propose a data-based non-parametric approach for identifiability analysis that is based on the bootstrap. It applies Dedicated session 2-2: Modelling the alternating conditional expectation algorithm to estimate so- called optimal transformations. Statistical analysis of the optimal approaches transformations allows for identifiability analysis regardless of model size or complexity. The algorithm identifies dependent, i.e. DS2-2-01 non-identifiable, groups of parameters, as well as the identifiable ones. We examplify the proposed procedure by applications to Construction of biochemical pathway models has to cope dynamical models of cellular signalling pathways. with nested uncertainties Conclusions: We show how identifiability analysis supports the Klipp, Edda1; Schaber, Jörg2; Liebermeister, Wolfram2 iterative cycle between modelling and experimentation in Systems 1Max Planck Institute for Molecular Genetics, Computational Biology. Systems Biology, Berlin, Germany; 2Max Planck Institute for Molecular Genetics, Berlin, Germany DS2-2-03

Objective: Traditional modeling approaches in computational Thermodynamic-kinetic modeling of complex formation in biology combine intuition about cellular regulation with selective signal transduction knowledge about the underlying biological processes. With the Ederer, Michael; Gilles, Ernst Dieter increasing accumulation of data concerning the stoichiometric Max Planck Institute for Dynamics of Complex Technical Systems, structure of biochemical networks and details of interaction and Magdeburg, Germany dynamics, grows the temptation to automatize the development of computational models. Still the problem remains that the Objective: The Wegscheider conditions follow from the principle available information is uncertain, incomplete or may not be of detailed balance and the second law of thermodynamics. They applicable to the specific situation considered such as cell type or constrain possible values of the kinetic parameters in reaction experimental scenario. networks. A mathematical model that violates these conditions Results: We discuss different approaches to automatically describes a thermodynamically impossible system. Large generating kinetic models of biochemical networks. One reaction networks contain usually a large number of Wegscheider strategy is the following: First, to characterize the network conditions. This makes the thermodynamically consistent, kinetic by its appropriate stoichiometric structure, since different modeling of such networks difficult. For this reason we developed stoichiometries might be compatible with the data. Second, the Thermodynamic-Kinetic Modeling (TKM) formalism, that to associate each reaction a kinetic equation. We either use provides a structurally consistent parameterization of kinetic conventional kinetics or convenience kinetics1, which is universally models (Ederer & Gilles, Biophys J, 92(6), 2007). The parameters applicable to any reaction with arbitrary numbers of reactants. of the TKM approach are capacities of species and resistances Last, the kinetics of the reactions of the networks are assigned of reactions. Networks that model the formation of multi-protein parameters. Since each level of assignment is dependent on the complexes contain a particularly high number of Wegscheider other ones and a full model must describe network, kinetics and conditions. Since the formation of protein complexes is a central parameters appropriately, one may also combine all three steps motif of cellular signal transduction, we will show how the TKM and only compare the fully characterized model against data. formalism can be applied to such networks. With respect to uncertainty on the parameter level, we discuss Results: The thermokinetic capacities of protein complexes

ICSB 2008 25 can be written by a product of a base capacity with interaction DS2-2-05 factors. The interaction factors depend on the binding energies of the proteins. Higher order interaction factors are possible, if An optimal identification procedure for model development more than two proteins interact. In a similar way, thermokinetic in systems biology resistances can be decomposed into a base resistance and Balsa-Canto, Eva; Alonso, Antonio A.; Banga, Julio R. interaction factors that describe the influence of the binding state IIM-Spanish Council Scientific Research, Process Engineering of the participating proteins. This decomposition of capacities and Group, Vigo, Spain resistances allows us to formulate a thermokinetic model in a rule- based manner. The nonlinear character and the usually large number of Conclusions: Kinetic models that describe the formation of parameters in biological mathematical models make model protein complexes are prone to a violation of the Wegscheider identification from experimental data a rather complex task. conditions. The Thermodynamic-Kinetic Modeling (TKM) The origin of such complexity may be threefold: the lack of formalism provides a thermodynamically consistent and intuitive structural identifiability, the lack of practical identifiability or both. parameterization of such networks by capacitive and resistive The structural identifiability problem is related to the pair model- interaction factors. It allows for a rule-based formulation of observable(s) whereas the practical identifiability is often related to Dedicated thermokinetic models. This is important since due to the the experimental scheme and experimental noise. Additionally the

Orals combinatorial complexity of protein complexation, the number of presence of several suboptimal solutions may be regarded also as occurring species and reactions increases exponentially with the a practical identifiability problem. number of proteins. Objective: This work proposes an iterative identification procedure to detect and deal with those difficulties. The DS2-2-04 procedure involves the following steps: first, usually disregarded, performing a structural identifiability analysisto detect which, if Analysis of biological network data using likelihood and not all, are the identifiable parameters and which are the best likelihood-free inference techniques candidates for identification; second,optimal experimental Wiuf, Carsten1; Oliver, Ratmann2; Knudsen, Michael1; Stumpf, design allows the computation of that scheme of experiments Michael3; Richardson, Sylvia3 that maximizes the quality and quantity of information for model 1University of Arhus, Aarhus, Denmark; 2Imperial College, London, calibration; third, the model calibration using global optimization United Kingdom; 3Imperial College, London, United Kingdom methods and fourth, a robust practical identifiability analysis to evaluate the quality of the parameter estimates. Objective: Biological Networks have received much attention Results: The procedure was successfully applied to several in recent years, but statistical tools for network analysis and examples related to the modelling of the MAP Kinase and NFKB modelling are still in their infancy. In this talk we focus on Protein signalling cascades, which are frequently involved in larger cell Interaction Networks (PINs) that typically comprise thousands of signalling pathways, and it is known to regulate several important proteins and interactions. A PIN is an empirical observation of an cellular processes. The methodology showed that it is impossible organism’s interactome which is the whole set of proteins and to simultaneously estimate all parameters in both examples and protein-protein interactions in the organism. that by the use of the proposed iterative procedure the final PINs are notoriously noisy due to experimental conditions, but identifiability properties are substantially improved with maximum they bear evidence of how evolution has shaped the interactome, expected uncertainties on the parameters around the order of what evolutionary processes are important and whether the same magnitude of the experimental error. processes are important in different species. Conclusions: The examples clearly revealed the usefulness of Results: In this study we adopt simple mathematical models the proposed identification procedure to improve efficiency and that capture the essentials of protein evolution and we develop robustness during model development in systems biology and statistical methods to estimate evolutionary PIN parameters. Our contributing to enhance model predictive capabilities. initial approach is based on Importance Sampling and a recursion for the likelihood, but this approach becomes computationally DS2-2-06 intractable for reasonably sized networks. Our second approach is based on likelihood-free inference, MCMC and summary Evolutionary simulation of cell differentiation under statistics. We discuss problems with selection of summaries, phenotypic noise convergence of MCMC, and credibility and apply the methods Tachikawa, Masashi on H. pylori and P. falciparum data. The results are interpreted Japan Science and Technology Agency, ERATO, Complex biologically. Systems Biology Project, Tokyo, Japan In addition we present mathematical results that back and explain some of the statistical problems we encounter when analysing Objective: How does the phenotypic noise affect evolution of PINs, e.g. when inferring the interactome size (number of links the development? Phenotypic noise caused by the stochasticity in the interactome). The mathematical models also allow us to in cellular processes affects the development, which can give learn about the dynamics of the evolutionary processes: Which the large phenotype variety even for a genetically homogeneous parameter choices stabilize network properties (e.g. the degree population. Although the evolution seems to be independent of distribution) over time and for which choices does stabilization not a phenotypic noise or phenotype variety at the first glance, some occur? The answer to this mathematical question will be related relations between them have been pointed out, e.g. canalization to the results we obtain from the analysis of the PIN data. and genetic assimilation proposed by Waddington. However, Conclusions: We have shown how inference about the evolution unified theory is still missing. of PINs are possible using advanced statistical techniques, but Results: To investigate the impact of phenotypic noise on the also that care should be taken in interpretation of the results - evolution of of the development, we performed evolutionary data is noisy and models might be improved. simulations. We present several gene regulatory network models for developmental dynamics of a cell group, and design the evolutionry tasks to acquire the spontaneous cell differentiation as the developmental dynamics. We found that the evolutionary processes of all models are clearly classified into two types, one is accelerated by the increase of phenotypic noise, the other is not. We also show that the bifurcation theory explains the difference between two types of evolution. The types of the bifurcation structures involved in these evolutionary processes determine the dependence on phenotypic noise. The structures are determined by the evolutionary tasks and independent from the size of the gene networks or other details of the models.

26 ICSB 2008 Conclusions: Since the bifurcation structures in dynamical model, an appropriate structure and mathematical expressions systems are mathematically well defined and are applied should be chosen, which most certainly will evolve during the systematically, our formulation provides an universal law for the model building process. The objective of image processing evolution of differentiation in multicellular systems. is to lump the spatiotemporal data content in the microscopy images into time-series, where the resulting aggregated entities DS2-2-07 reflect something that is described by the model or vice versa. Estimation of model parameters is performed both on individual Kinetic modeling and analysis of nonlinear biochemical and population level using a maximum likelihood approach. networks with no quantitative information Certain aspects of this framework are exemplified by results from Musters, Mark1; de Jong, Hidde2; van den Bosch, Paul3; van Riel, an investigation on the RAS/cAMP/PKA-pathway in yeast. Natal4 Conclusions: The framework provides an integrated approach 1Wageningen University, Laboratory of Microbiology, Wageningen, for the process of system identification using microscopy images Netherlands; 2INRIA Rhone-Alpes, Saint-Ismier, France; of cell populations, and puts emphasis on an iterative workflow. 3Eindhoven University of Technology, Department of Electrical It also highlights, and aims at quantifying, an important fact of Engineering, Eindhoven, Netherlands; 4Eindhoven University of biological systems: the inter-individual variability that suggests that Technology, Department of Biomedical Engineering, Eindhoven, many parameters in single cell models should not be thought of Netherlands as fixed but are better understood from their statistical properties in a population perspective.

Objective: The complexity of biochemical networks is enormous Orals and difficult to unravel by intuitive reasoning alone. Kinetic Dedicated session 2-3: Diagnostic Dedicated modeling has traditionally been proposed as tool for the analysis of network dynamics. However, one of the major bottlenecks of markers and complex diseases computational modeling is the lack of quantitative information, which is a necessity for simulation and system identification. The DS2-3-01 objective of this study was therefore to develop a method that can simulate and analyze the dynamical behavior of nonlinear Clinical systems biology to personalize medication biochemical networks without requiring accurate time-series data, Benson, Mikael reliable parameter values or other quantitative data. To validate Unit for Clinical Systems Biology, Göteborg, Sweden the practical relevance of our approach, a nonlinear kinetic model of extracellular matrix (ECM) remodeling was chosen. Common diseases like allergy, obesity and cancer are complex. Results: We developed a qualitative modeling procedure that is Each of these diseases is caused by altered interactions between able to link various types of dynamical behavior (limit cycle, single multiple genes. These alterations may differ between different steady state, multiple steady states) of the complex ECM model individuals although they appear to have the same disease. A to unique sets of parameter inequalities. The outcome of this clinical consequence is that a medication that works for one qualitative analysis was similar to results found by cumbersome patient does not work for another. In some individuals the numerical analysis of the original model. medication may even cause severe side-effects and deterioration Conclusions: In our test case, qualitative analysis deduces three of the disease. Ideally, physicians should be able to personalize sets of inequalities, constraining the parameters to guarantee medication based on simple tests that measure a few proteins bistability in the ECM model. When these parameter restrictions in saliva or blood. We aim to identify such proteins using high- are not satisfied, only a single stable steady state is present. throughput technology, high-performance computing and These observations correspond exactly to extensive numerical systems biology. exploration of the phase space. Common hay fever is used as a model of complex disease. This The scarcity of quantitative information and large amounts of is because we have a well-defined disease model that can be qualitative data in current biological research have motivated the studied with out risk or discomfort in a large number of patients. development of a qualitative modeling method that sufficiently White blood cells are challenged with allergen in vitro and mRNA accurate describes and predicts the dynamics of nonlinear expression of all human genes examined with DNA microarrays. biochemical networks. Our approach can be applied to a large Network-based analysis is performed to find transcriptomal sub- variety of biochemical networks and might assist in smart networks that are specific for different disease sub-types. experimental design, parameter estimation, metabolic engineering The sub-networks are dissected to find pathways and regulatory and identifying specific drug targets. genes. These genes are examined for disease-associated polymorphisms and if their corresponding proteins can be used DS2-2-08 as diagnostic markers. The project is based on a multi-disciplinary collaboration between twelve research groups in Europe and the System identification from spatiotemporal cell population US that is funded by two EU grants. data Almquist, Joachim; Sunnåker, Mikael; Hagmar, Jonas; DS2-3-02 Kvarnström, Mats; Jirstrand, Mats Fraunhofer-Chalmers Centre, Göteborg, Sweden Understanding networks of atherosclerosis Tegnér, Jesper; Björkegren, Johan; Skogsberg, Josefin; Maleki, Objective: Fluorescence microscopy is a powerful technique for Shohreh; Noori, Peri in vivo visualization of localization processes, levels of expression, Department of Medicine, Stockholm, Sweden protein kinetics, and protein-protein interactions, at the level of individual cells. When studying cell cultures, the population aspect Within the systems biology community there has been an adds another dimension to the available information for parameter increasing interest in employing a systems approach targeting estimation, compared to single cells. A single experiment medical problems including complex diseases. In parallel there is generates an abundance of data and it is necessary to be able to a rapid development of genomic technologies and the medical handle uncertainties for individual cells as well as the variations research community has begun to appreciate the need for between cells. As a result, one is faced with a hierarchical computational tools in order to deal with the complexity of the system identification problem where single cell models have to data. I will talk about our (www.compmed.se) systems biology be extended with an additional level describing the variation of approach targeting atherosclerosis which is a major killer in parameters between individuals of the population. the western world. We integrate samples (tissues, blood) from Results: We present a system identification framework including well characterized by-pass surgery patients, a conditional LDL single cell modeling, image processing, and parameter estimation. mouse model for atherosclerosis and siRNA experiments using The framework is intended for dynamic models with both intrinsic THP-1 cells as models for macrophages. By thus approach we and extrinsic stochastic elements. Given the purpose of the have obtained the first draft of the core gene network driving the

ICSB 2008 27 macrophage related plaque transformation in atherosclerosis. role in tumorigenesis of human and rodent. We have now finished the proof-of-concept phase and are now Conclusion: The cross-species comparative oncogenomic in the process of expanding our systems approach increasing the analysis could effectively filter putative cancer genes from number of patients and the number of data-types including SNP heterogeneous and complicated oncogenomic data. Furthermore, studies. this strategy would be valuable in dissecting the HCC tumorigenesis pathway between human and rodent as well as DS2-3-03 identifying representative rodent HCC models which would have plenty application in the HCC diagnosis and therapy. Ultra short tandem repeats: A target for disease mutations Eskerod Madsen, Bo; Villesen, Palle; Wiuf, Carsten DS2-3-05 University of Aarhus, BiRC, Aarhus, Denmark Transcriptional variations associated with Parkinson’s Objective: In recent years it has been demonstrated that disease structural variation, such as indels (insertions and deletions), Orr-Urtreger, Avi1; Bar-Shira, Anat2; Giladi, Nir3; Kedmi, Merav2 are common throughout the genome, but the implications of 1Tel Aviv Sourasky Medical Center, Genetic Institute & Sackler Dedicated structural variations are still less clearly understood. Long tandem Faculty of Medicine, Tel Aviv, Israel; 2Tel Aviv Sourasky Medical

Orals repeats (e.g. microsatellites or simple repeats) are known to Center, Genetic Institute, Tel Aviv, Israel; 3Tel Aviv Sourasky be hypermutable (indel-rich), but are rare in exons and only Medical Center, Department Neurology & Sackler Faculty of occasionally associated with diseases. Here we focus on ultra Medicine, Tel Aviv, Israel short, imperfect tandem repeats (USTRs) and investigate whether they share the hypermutability of the longer tandem repeats and Objectives: Parkinson’s disease (PD) is the second most whether disease-related genes have a higher USTR content than common neurodegenerative disorder, affects 1-4% of the non-disease-related genes; in short we aim at testing whether worldwide population over 60 years. While environmental USTRs are targets for disease-related mutations in human exons. factors play a role in disease development, the contribution of Results: We show that validated human indels are extremely inherited genetic variations is considerably higher than previously common in USTR regions compared to non-USTR regions. In appreciated. The unique structure and origin of the relatively contrast to longer tandem repeats, we found that USTRs are homogeneous Ashkenazi population were valuable for studying present in exons of most known human genes (92.23%), 99% carcinogenesis and genetic syndromes, and founder mutations of all USTR sequences in exons are shorter than 33 base pairs detected in Ashkenazim were identified in genes associated with and 62.1% of all USTR sequences are imperfect repeats. We diseases in the population-at-large. Thus, studying this population also demonstrate that USTRs are significantly overrepresented in is also expected to shed a light on genetic factors underlying disease-related genes in both human and mouse. These results PD pathogenesis. We therefore established the largest cohort are preserved when we limit the analysis to USTRs outside known of Ashkenazi PD patients and detected mutations in the LRRK2 longer tandem repeats. or GBA genes in a surprisingly high proportion, of about a third, Conclusions: Based on our findings we conclude that USTRs of the patients. To detect novel genes and genetic pathways represent hypermutable regions in the human genome that associated with PD, we determined the transcriptional profiles in are linked to human disease. In addition, USTRs constitute an patients with and without mutations. obvious target when screening for rare mutations, because of the Results: The analysis included 119 RNA samples derived relatively low amount of USTRs in exons (1,973,844 bp in human; from peripheral blood leukocytes (PBL) of PD patients with and 1,544,242 bp in mouse) and the limited length of USTR regions. without mutations in the LRRK2 or GBA genes, and age-matched controls, and was done using the Affymetrix GeneChip Human DS2-3-04 Exon 1.0 ST Arrays. Bioinformatics performed on the “core data” (21,980 genes and 232,448 exons) revealed expression changes Discovery of cancer genes through comparative on gene-, transcript variant- and exon-levels. The transcriptional oncogenomic analysis of hepatocellular carcinoma alterations were examined between PD patients and controls and Su, Wen-Hui1; Ho, Chun-Ming1; Jou, Yuh-Shan2 between the different genotype groups of PD patients. Alternative 1Chang Gung University, Molecular Medicine Research Center, splice ANOVA detected 106 differentially expressed transcript Taoyuan, Taiwan; 2Academia Sinica, Institute of Biomedical variants between PD patients and controls, and 139 between the Sciences, Taipei, Taiwan 4 experimental groups (Bonferroni corrected p<0.05). Conclusions: Our results suggest that this study population is a Objective: Recently, several studies using cross-species valuable tool for studying the complex genetic basis of PD, and comparative genomic analysis has proved to be a powerful that PBL may serve as a relevant surrogate tissue to examine the strategy to identify putative cancer genes. The alignment of transcriptional changes involved in PD pathogenesis and risk. chromosome aberration and the consistence of gene expression between mouse and human tumors could effective narrowed DS2-3-06 down the minimum overlapping region and help to identify putative cancer genes. Previously, we have constructed an open Placing disease mutations in context: Chemical genomics access database OncoDB.HCC (http://oncodb.hcc.ibms.sinica. and synthetic interaction screens in human cells edu.tw) which integrated various genomic data in HCC to provide Shaw, Stanley1; Subramanian, Aravind2; Westly, Elizabeth2; Ma, lines of evidence for revealing important aberrant cancer target Maggie1; Blodgett, David1; Schreiber, Stuart2 genes and loci. 1Massachusetts General Hospital, Center for Systems Biology, Results: Via comparative mapping of HCC loci in between Boston, United States; 2Broad Institute of Harvard and MIT, human aberrant chromosomal regions and quantitative trait loci Cambridge, United States (QTLs) of mouse and rat HCC models we revealed 12 syntenic HCC regions and the putative human cancer genes located in Objective: Complex diseases such as type 2 diabetes arise those regions. In this study, the comparative oncognomic analysis from the interaction between multiple genes, environment and was performed by comparing the expression profiles between behavior. Synthetic interaction screens have revealed how orthologous genes and then combined with the syntenic HCC pathways intersect in genetically tractable organisms. Here, we region data. The concordantly expressed genes in between employ a chemical genomic approach to perform a synthetic human and rodent HCCs were analyzed by gene ontology interaction screen in patient-derived cells, in order to identify and pathway analysis. We further integrated those concordant pathways that interact with disease mutations. As proof of expressed genes onto 12 syntenic HCC regions and mouse/rat concept, we studied lymphoblastoid cell lines (LCLs) from families QTL for identification of putative cancer genes. Our results also with an extreme form of diabetes (Maturity Onset Diabetes of the suggested those genes could be the potential cancer genes for Young type 1, or MODY1) caused by mutations in the orphan the corresponding rodent HCC models and played an important nuclear receptor HNF4α. LCLs have been created as part of

28 ICSB 2008 many patient cohorts, and are increasingly being explored as otherwise well-treated CAD. A new transcription factor may be a model system for genetic studies. Results: 24 LCLs from a high-hierarchy regulator of this pathway and thus a potential family members that are mutant or wild-type at HNF4α were therapeutic target. screened with approximately 4,000 small molecules of known effect, including several hundred drugs; small molecule effects DS2-3-08 were assessed using high-throughput metabolic viability assays. We identified several small molecules that showed quantitatively Identification of an IRF4 regulated module by combined distinct phenotypes in mutant vs. wild-type cells, including several ChIP-chip and gene expression analysis glucocorticoids in clinical use, as well as certain classes of fatty Mobini, Reza1; Bengt A., Andersson2; Cardell, Lars Olaf3; Hahn- acids. This suggests that these small molecules (or associated Zoric, Mirjana2; Langston, Michael A.4; Perkins, Andy D.4; Rak, proteins) functionally interact with HNF4α. We also performed Sabina5; Benson, Mikael1 genome-wide gene expression studies on LCLs and identified 1University of Gothenburg, The Unit for Clinical Systems Biology, metabolic pathways whose activity is altered in HNF4α-mutant Gothenburg, Sweden; 2Sahlgrenska University Hospital, Clinical cells. Together, these data suggest testable hypotheses about the Immunology and Transfusion Medicine, Gothenburg, Sweden; disease mechanisms underlying MODY1 diabetes. 3Karolinska Institute, Department of Oto-Rhino-Laryngology, Conclusions: This approach applies the logic of genetic Gothenburg, Sweden; 4University of Tennessee, Electrical synthetic interaction screens to study pathway interactions in Engineering and Computer Sciences, Knoxville, United States; patient cells, and may help assign function to disease mutations 5Sahlgrenska University Hospital, Respiratory Medicine and

of unknown mechanism. Furthermore, this represents an Allergy, Gothenburg, Sweden Orals unbiased approach to identify small molecules and pathways that Dedicated can distinguish, or classify, distinct cell states (such as diseased Objective: Identification of modules of functionally related vs. healthy, or mutant vs. wild-type). Finally, drugs and proteins genes may facilitate understanding of pathogenic mechanisms that interact with a disease mutation are potential modifiers of in complex diseases such as human allergic inflammation. disease, and may suggest novel therapeutic approaches. In this study we aim to examine the role of IRF4 in seasonal allergic rhinitis. We propose that combined ChIP-chip and gene DS2-3-07 expression array analysis may be more generally applicable to help find functionally related gene modules in complex diseases. Multi-organ expression profiling uncovers transendothelial Results: IRF4 was found to be the transcription factor that migration of leukocytes and a transcription factor as increased most in expression in allergen-challenged CD4+ cells. potential targets in coronary artery disease: The Stockholm Its relevance for proliferation of Th2 cells was indicated by positive Atherosclerosis Gene Expression (STAGE) study correlations with the Th2 cytokines IL-5 and IL-13. Treatment Hägg, Sara1; Skogsberg, Josefin1; Lundström, Jesper1; Noori, with Corticosteroids decreased IRF4, IL-5 and IL-13 significantly. Peri1; Nilsson, Roland2; Maleki, Shohreh1; Brinne, Björn2; IRF4- and RNA polymerase II chromatin-immunoprecipitation Bradshaw, Maria1; Bajic, Vlad3; Samnegård, Ann1; Silveira, promoter tiling arrays resulted in the identification of putatively Angela1; Gigante, Bruna1; Leander, Karin1; de Faire, Ulf1; IRF4-regulated genes. These genes were filtered by selecting Rosfors, Stefan1; Lockowandt, Ulf1; Liska, Jan1; Konrad, Peter1; those that changed in expression in allergen-challenged CD4+ Takolander, Rabbe1; Franco-Cereceda, Anders1; Ivert, Torbjörn1; cells and correlated with IRF4. This resulted in 296 genes that Hamsten, Anders1; Tegnér, Jesper2; Björkegren, Johan1 were IRF4 induced- and 91 that were IRF4 repressed. The IRF4 1Karolinska Institutet, Stockholm, Sweden; 2Linköping Institute induced genes included genes that specifically induce Th2 cell of Technology, Linköping, Sweden; 3South African National proliferation, such as GATA3, as well as general activators like Bioinformatics Institute, Cape Town, South Africa IL2RA and IL2RG. Increased expression of these genes was validated with real-time PCR. In contrast, the IRF4-repressed Objective: Stress imposed by organ-regulating energy genes included anti-inflammatory genes, like IL10. metabolism leads to arterial lipid accumulation and inflammation Conclusion: This pioneering study has demonstrated that a and eventually to coronary artery disease (CAD), the number module of co-regulated genes can be recognized in complex one killer in Westernized societies. To identify functional modules diseases by identifying a key transcription factor and then defined by gene expression central to atherogenesis, we analyzed exploiting a combination of gene expression arrays and ChIP-chip a compendium of expression profiles of multiple CAD-relevant analysis. Using this strategy we point to a key role for IRF4 in organs in humans. allergic inflammation by the induction of pro-inflammatory genes Results: In the Stockholm Atherosclerosis Gene Expression and the repression of anti-inflammatory genes. study, 114 well-characterized CAD patients undergoing coronary artery bypass surgery were enrolled over 3 years. Sixty-six Dedicated session 2-4: Microbial systems patients underwent full metabolic profiling (liver, skeletal muscle, and visceral fat). Gene expression profiles were also obtained from the atherosclerotic and unaffected arterial walls in 40 of DS2-4-01 these patients, resulting in a compendium of 278 gene expression profiles (n = 15,042 genes/profile). Superparamagnetic clustering From membrane proteins to their interactomes: Lessons of mRNA levels generated 60 gene expression modules (n = from yeast and humans 4007 genes) from all organs. Two modules related to the extent Stagljar, Igor of coronary stenosis, one in aortic lesions (n = 49) and one in University of Toronto, Toronto, Canada mediastinal fat (n = 59). Remarkably, in a validation cohort of 25 carotid stenosis patients, 27 of these genes were replicated in the Due to their pivotal role in many cellular processes, their direct link only module (n = 55), out of eight, that related to carotid stenosis. to human diseases and their extracellular accessibility to drugs, In all three modules, genes belonging to the transendothelial the identification of proteins associated with integral membrane migration of leukocyte pathway were clearly overrepresented. proteins is desirable. However, due to their complex biochemical A transcription factor expressed in lesion macrophages and properties, membrane proteins are very hard to manipulate, endothelial cells was identified as a potential high-hierarchy making the study of their corresponding interactors even more regulator of this pathway. The minor T-allele of a single nucleotide challenging. polymorphism in the transcription factor was inversely related Previously, our lab developed a yeast-based genetic technology to mRNA levels and extent of coronary atherosclerosis and was for the in vivo detection of membrane protein interactions, called underrepresented in myocardial infarction survivors versus healthy the split-ubiquitin membrane yeast two-hybrid (MYTH) system. controls; these findings were replicated in two independent CAD Our current efforts are directed to identify, characterize, and cohorts. perturb focused collections of membrane proteins in an effort to Conclusions: Transendothelial migration of leukocytes understand complex biological processes such as cell signaling appears to be rate limiting in lesion development in severe but and membrane transport at a systems level. During my talk, I will

ICSB 2008 29 discuss exciting new findings indicating that the newly identified DS2-4-03 interactors play novel roles in regulating the activity of several yeast ABC transporters, and human G-protein coupled receptors A genomic approach to map transcription factor and (GPCRs), receptor tyrosine kinases and transporters. Our initial kinase pathways in budding yeast success suggests that the MYTH system, whose unique strength Andrews, Brenda1; Kainth, Pinay2; Sharifpoor, Sara2; van Dyk, lies in its ability to identify interactors for the full-length integral Dewald2; Sassi, Holly2; Pena-Castillo, Lourde2; Kostic, Alex2; membrane proteins, represents a robust technology that will be Boone, Charles2; Hughes, Timothy2 versatile in identifying key interactors for the majority of integral 1University of Toronto, Donnelly CCBR, Toronto, Canada; membrane proteins from any organism. 2University of Toronto, Banting & Best Dept of Medical Research, Key references: Toronto, Canada Stagljar, I., Korostensky, C., Johnsson, N., and te Heesen, S. (1998) Proc Natl Acad Sci USA 95, 5187-5192. Objective: We are using tools of yeast functional genomics to Thaminy, S., Auerbach, D., Arnolodo, A., and Stagljar, I. (2003) systematically define pathways controlling conserved biological Genome Res 13, 1744-1753. processes in Saccharomyces cerevisiae. Our general strategy Paumi, C.M., Menendez, J., Arnoldo, A., Engels, K., Iyer, K., uses an automated genetics platform called synthetic genetic Dedicated Thaminy, S., Georgiev, O., Barral, Y., Michaelis, S., and Stagljar, I. array (SGA) analysis to enable the high throughput examination

Orals (2007) Mol Cell 26, 15-25. of the effects of genetic perturbations on the viability of kinase Gisler, S.M., Kittanakom, S., Fuster, D., Radanovic, T., Wong, V., mutants or on reporter gene expression. Bertic, M., Hall, R.A., Engels, K., Murer, H., Biber, J., Markovic, Results: Our transcription factor project involves introduction of D., Moe, O.W., and Stagljar, I (2008) Mol Cell Proteomics, Apr 11, promoter-GFP reporter constructs along with a control promoter- 2008; [Epub ahead of print]. RFP gene into the array of ~5000 yeast deletion mutants. Fluorescence intensities from each reporter are assayed from DS2-4-02 individual colonies arrayed on solid agar plates using a scanning fluorimager and the ratio of GFP to RFP intensity reveals deletion Quantitative morphological traits of budding yeast is mutants that cause differential GFP expression. So far, we have applicable to identify drug targets used this approach to examine expression of a panel of 27 cell Ohya, Yoshikazu1; Oka, Satomi2; Ohnuki, Shinsuke2; Nogami, cycle regulated promoter-GFP constructs that represent all cell Satoru2 cycle phases and used our reporter system to discover new 1University of Tokyo, Integrated Biosciences, Kashiwa, Japan; regulators of these genes. By doing so, we created a data matrix 2Integrated Biosciences, Kashiwa, Japan of quantitative gene expression measurements representing 27 reporter genes by 5000 yeast deletion mutants. In our screening Objective: Yeast cell morphology is an attractive target for effort, we have identified both known and novel candidate comprehensive analysis, because it reflects various cellular regulators of cell cycle genes. For our kinase project we are events, including progression through the cell cycle, establishment using systematic Synthetic Dosage Lethality (SDL) screening as a of cell polarity, and regulation of cell size control. We recently powerful genetic tool to identify kinase substrates. SDL is based developed an image-processing system that automatically on the idea that increasing levels of a protein may not affect the processes digital images of over two hundred budding yeast growth of a wild-type strain, but may cause a clear phenotype cells to obtain quantitative morphological data of yeast mutant (e.g. lethality) in strains disrupted for pathway components or strains. Our high-dimensional and quantitative phenotyping of interacting proteins. We have now performed SDL genome-wide 4,718 haploid systematic deletion mutants revealed that similar screens on all non-essential yeast kinases in order to identify phenotypes are caused by deletions of functionally related novel substrates and to better understand the biology of yeast genes, enabling a functional assignment of a locus to a specific kinases. cellular pathway. Here, by using this morphological database, we Conclusions: We anticipate that our systematic screens will propose a novel strategy to identify the target proteins of drugs provide the first comprehensive map describing the interplay of by searching the similar morphological phenotype after the drug regulators controlling the eukaryotic cell cycle and a complete treatment. roster of potential kinase targets. Results: We tested the validity of the scheme with well- characterized drugs such as hydroxyurea, a non-alkylating DS2-4-04 myelosuppressive agent that inhibits DNA synthesis and latrunculin A, an inhibitor of the assembly step in a rapid cycle MAPK-dependent negative feedback in pheromone of binding to monomeric G-actin. Morphological changes of the response regulates information transmission drug-treated cells were monitored after staining mannoprotein Yu, Richard1; Pesce, C. Gustavo1; Colman-Lerner, Alejandro2; (as a cell wall component marker), the actin cytoskeleton and Lok, Larry1; Pincus, David3; Serra, Eduard4; Holl, Mark5; Benjamin, nuclear DNA. We found that morphological phenotypes of the Kirsten6; Gordon, Andrew7; Brent, Roger1 drug-treated cells resemble those of the mutant cells defective 1The Molecular Sciences Institute, Berkeley, CA, United States; in their target proteins. The results imply that it is possible to 2Instituto de Fisiología, Biología Molecular y Neurociencias, identify biological targets for drug action using comprehensive CONICET and Universidad de Buenos Aires,, Fisiología, Biología morphological database of the yeast deletion mutants. Molecular y Celular, Buenos Aires, Argentina; 3University of Conclusion: This novel method is theoretically applicable California, San Francisco, San Francisco, CA, United States; to identify any gene products that functionally interact with 4Institut d’Investigació Biomèdica de Bellvitge (IDIBELL), Centre small molecules, and therefore will generally provide powerful de Genètica Mèdica i Molecular, Barcelona, Spain; 5Arizona State techniques for inferring mechanism of the drug action. University, Biodesign Institute, Tempe, AZ, United States; 6Amyris Biotechnologies, Emeryville, CA, United States; 7Brookhaven National Laboratory, Physics, Upton, NY, United States

Objective: Determine mechanisms regulating information transmission in signal transduction pathways. Results: We identified a rapid non-transcriptional negative feedback in the yeast pheromone response system, mediated by the MAPK Fus3, that improves the ability of cells to measure and respond to different doses of pheromone. We propose that the feedback, by making the linear ranges of receptor-pheromone binding response and downstream responses match, increases both the range of doses over which the cells can respond and the precision of their responses (i.e., increases the amount of

30 ICSB 2008 information the system can transmit). nucleoid, mitochondrial morphology, and apoptosis. We show Conclusions: Our work reveals a new molecular mechanism that yeast merge a number of cellular processes, which we call for regulating signal intensity in cell signaling systems, and modules, into a longevity network. Implementing our knowledge, shows how concepts from pharmacology, noise and variation we developed a high-throughput assay that was used to identify in signaling, and information theory can help us study and novel anti-aging small molecules and to establish their cellular understand how signaling systems transmit information. targets. Conclusions: Our findings suggest a model for the DS2-4-05 spatiotemporal dynamics of the longevity network in yeast. This model envisions that 1) yeast establish a diet- and genotype- Drug interactions modulate the potential for evolution of specific configuration of the network by setting up the rates resistance of the processes taking place within each of its modules; 2) Michel, Jean-Baptiste1; Yeh, Pamela1; Kishony, Roy1; Moellering, the establishment of a network’s configuration occurs before Robert2 yeast enter a non-proliferative state; and 3) different network’s 1Harvard Medical School, Systems Biology, Boston, United configurations established prior to entry into a non-proliferative States; 2Beth Israel Deaconess Medical Center, Boston, United state define different rates of survival following such entry. Thus, States by designing a specific configuration of the modular longevity network prior to reproductive maturation, yeast define their life Objective: Antimicrobial treatments increasingly rely on multi- span. Using our understanding of the longevity network, we

drug combinations, due in part to the emergence and spread identified novel anti-aging small molecules that target the lipid Orals of resistance to single antibiotics. The continued effectiveness metabolism hub of the network. Dedicated of combination treatments depends crucially on the frequency at which multi-drug resistance arises. Yet, it is unknown how DS2-4-07 this frequency is affected by cross-resistance and epistatic interactions—ranging from synergy to antagonism—between Identification of minimal net stoichiometries and minimal the drugs. Here we analyze how interactions between drugs substrate sets in metabolic models affect the spontaneous emergence of resistance in the medically Gevorgyan, Albert1; Poolman, Mark1; Anthony, Mark2; Fell, David1 important pathogen Staphylococcus aureus. 1Oxford Brookes University, School of Life Sciences, Oxford, Results: Introducing a new experimental method for high- United Kingdom; 2Birmingham Women’s Health Care Trust, throughput colony imaging, we counted resistant colonies arising Birmingham, United Kingdom across a 11x11 matrix of drug concentrations for each of three drug pairs. Resistance is selected for within a window of drug Objective: Elementary modes analysis (Schuster et al, 99) concentrations high enough to inhibit wild-type growth, but low provides a comprehensive description of a biochemical system enough to allow some resistant mutants to grow. Our data show by detecting all minimal reaction pathways which can operate that the different drug combinations have significantly different at a steady state. However, genome-scale models may contain impacts on the size of this window of concentrations which millions of elementary modes and their calculation is often favors the evolutionary adaptation of resistance. We frame these computationally infeasible. Elementary modes are characterised results in a novel mathematical model in which the frequencies by net stoichiometries, which represent interconversions of of resistance to single drugs, cross-resistance, and epistasis external substrates and products; multiple elementary modes may combine to determine the propensity for multi-drug resistance. have the same net stoichiometry. The knowledge of the possible Conclusions: The theory suggests that drug pairs which interact net stoichiometries of a network enables the identification of such synergistically, preferred for their immediate efficacy, may in fact practically important properties, as the maximal product yield by a favor the future evolution of resistance. This framework reveals unit of substrate and the composition of minimal media. the central role of drug epistasis in the evolution of resistance and Results: We introduce the concept of minimal net points to new strategies for combating the emergence of drug- stoichiometries: structural invariants of metabolic networks resistant bacteria. describing the elementary net conversions of external metabolites. We also define minimal substrate sets as irreducible sets of DS2-4-06 nutrients which are sufficient for the production of a given external product, e.g protein. Algorithms for the identification of both The spatiotemporal dynamics of a modular network that network properties are proposed. Their application to genome- regulates longevity in yeast and is controlled by a novel scale metabolic models of S. agalactiae (constructed in-house) class of anti-aging small molecules and E. coli (Reed et al 03) is presented. In both models, amino Titorenko, Vladimir; Goldberg, Alexander; Gregg, Christopher; acids and protein were defined as the only external metabolites. Boukh-Viner, Tatiana; Kyryakov, Pavlo; Bourque, Simon; Aziz, 23 and 8455 minimal net interconversions of amino acids were Zeinab; Chang, Andrew; Cyr, David; Kayembe, Mulanda; Kim, found in the models of S. agalactiae and E. coli, respectively. Hyun Young; Machkalyan, Gayane; Milijevic, Svetlana; Mudhar, In S. agalactiae, 8 minimal combinations of amino acids were Ramandeep; Quashie, Peter; Ramlal, Nishant; Uscatescu, Victor; identified which are sufficient for protein biosynthesis; each of Askari, Mohammad them consisting of 14 amino acids, 12 out of which are essential. Concordia University, Biology Department, Montreal, Canada In E. coli, 79 such combinations were found, consisting of 2 or 3 amino acids. Objective: The fundamental mechanisms of aging are conserved Conclusions: The calculation of minimal net stoichiometries is across phyla. The yeast Saccharomyces cerevisiae is a feasible in large genome scale models, such as the model of E. valuable model for unveiling the mechanisms of cellular aging coli. Identification of minimal substrate sets is a useful method for in multicellular eukaryotes. Yeast aging can be slowed down by the analysis of nutritional requirements of the modelled organisms. calorie restriction (CR), a low-calorie dietary regimen that extends life span and delays age-related disorders in a wide spectrum of DS2-4-08 organisms. We sought to define a specific pattern of metabolism and organelle dynamics that is responsible for the anti-aging Longitudinal study of pseudomonas aeruginosa isolates effect of CR and to establish the mechanisms underlying such from Cystic Fibrosis lungs effect. Yang, Lei Results: We assessed the effect of a CR diet and numerous Technical Univestity of Denmark, Lyngby, Denmark mutations extending life span on the age-dependent dynamics of cellular and organellar proteomes and lipidomes, Objective: Cystic Fibrosis (CF) infection by the opportunistic carbohydrate and lipid metabolism, interorganellar metabolic bacterial pathogen Pseudomonas aeruginosa is an example of flow, concentration of reactive oxygen species, frequencies long-term persistent infections. In contrast to acute infections, of nuclear and mitochondrial DNA mutations, mitochondrial which damage the host by virulence factors, chronic infections

ICSB 2008 31 involve considerable genetic adaptation of the pathogens we studied how budding yeast integrates information when two during their coexistence with the host. However, little is known signaling systems are stimulated at the same time. We chose about the evolutionary strategies employed. Here we focused the mating pheromone response and the high osmolarity (HOG) on the changes in the transcriptome during the adaptation of pathways, which share several components, and measured P. aeruginosa using Affymetrix GeneChip microarrays. Whole the activity of each pathway using cells expressing specific genome sequencing and series of phenotypic assays were also fluorescent protein-based transcriptional reporters. used to study the bacterial evolution systematically. Results: The fundamentally different function of these two Results: We chose a CF patient from the Danish CF centre, pathways was reflected in the duration of their respective who has been chronically infected since 1991 by one particular responses. In the presence of continued stimulation, the response transmissible clone. The mucoid phenotype started to persist to pheromone was sustained, showing little to no adaptation from 2007. Seven P. aeruginosa isolates (5 non mucoid and 2 for at least three hours. On the other hand, the response to mucoid) from different stages of the infection were used in this high osmolarity was transient, as the response shuts itself off study with PAO1 (representing environmental strains) as the after causing an increase in internal osmolite concentration that reference. Compared to PAO1, the early isolates have many matches the increase in external osmolarity. Co-stimulation genes associated with antimicrobial susceptibility and LPS with high osmolarity and pheromone showed that cells respond Dedicated modification which are up-regulated, while genes related to to both stimuli with no apparent correlation in their level of

Orals motility and quorum sensing are down-regulated. Most of these activation in individual cells. Surprisingly, a 90 min pretreatment results have been confirmed by phenotype analysis. When with pheromone changed the high osmolarity transcriptional comparing late and early CF isolates, genes related to metabolism response from transient into sustained, for at least 3 hours. We are expressed most differently. found that this change required Mpk1, the MAP kinase of the cell Conclusions: The results indicate that a successful P. aeruginosa wall integrity pathway (CWIP), suggesting that Mpk1 modulates persistent strain has to go through a two-part adaptation. the high-osmolarity response during the pheromone response. The first one is the changes from the environment to a much In support of this interpretation, we found that activation of the more stressed condition. This allows the bacterium to avoid CWIP by high temperature also caused sustained HOG activation. host immune attacks and antibiotics. The second part is the Thus, the HOG pathway seems to be able to integrate inputs from optimization of metabolic pathways to best utilize the nutrients in other cellular pathways. Consistent with this idea, we observed the lung. Our model is the first systematic picture of P. aeruginosa higher cell to cell variation in the transcriptional output of the HOG infection in CF lungs. It has important consequences for designing pathway than in that of the pheromone pathway. new therapies of CF and understanding chronic microbial Conclusions: These results show how the normally transient infection in general. response of a signaling pathway can be turned into a sustained response by a modulating input from another pathway. We Dedicated session 3-1: hypothesize that the cell wall integrity pathway partly counteracts the high-osmolarity response, thereby forcing the continued Cell-to-cell variation activity of the Hog pathway.

DS3-1-01 DS3-1-03

Transcription factor search kinetics in bacterial cells Quantifying Hog1p MAPK signaling at the single cell level Elf, Johan1; Li, Gene-Wei2 Pelet, Serge; Rudolf, Fabian; Peter, Matthias 1Uppsala University, Cell and Molecular Biology, Uppsala, ETHZ, Institute of Biochemistry, Zürich, Switzerland Sweden; 2Harvard University, Cambridge, MA, United States Objective: Due to the stochastic nature of cellular systems, Objective: Motivated by our recent single molecule measurement uncovering regulatory mechanisms in signaling cascades requires of how fast a transcriptional repressor finds its specific binding quantitative measurements at the single cell level. We set out site in the bacterial genome (Elf et al. Science 2007) we have to study the Hog1 MAPK pathway in yeast as a model system adapted the facilitated diffusion theory (Berg et al. Biochemistry for stress activated protein kinase pathways in eukaryotes. To 1981 ) for target search by DNA binding proteins to the in vivo capture the behavior of individual cells, we developed assays situation. based on flow cytometry and fluorescent microscopy coupled Results: We identify two major differences as compared to the in to algorithms for automatic image analysis to quantify signaling vitro situation. The first difference is the high concentration of non- in yeast. Results: The HOG1-MAPK pathway regulates the specific DNA that sequesters the transcription factor and thus stress response to osmotic change of the environment. The slows down the search process. The second and more interesting relocation of the fluorescently-tagged MAP kinase Hog1p from difference is the high occupancy of other proteins on the bacterial the cytoplasm to the nucleus enables us to monitor the dynamics chromosome which restrict the open sliding distances on DNA of its phosphorylation and thus pathway activation in individual and also frequently blocks the specific binding site. We derive cells. Moreover, by quantifying the expression of a fluorescent analytical expressions for how the rate of binding to a specific site reporter construct we are able to follow the expression output of is reduced when these factors are considered and compare to the pathway. By combining these two assays in the same cell, we experimental results. can directly correlate the pathway capacity with the transcriptional Conclusions: A combined experimental and theoretical output in individual yeast cells. Interestingly, we uncovered a approach makes it possible to address the problem of how just discrepancy between the pathway and expression outputs of the a few transcription factor molecules per cell is enough to find a system. While the MAPK activation is relatively uniform across specific binding site sufficiently fast to accurately adapt to new a whole population of cells and increases gradually with stress, growth conditions or to maintain a high repression ratio. the protein expression shows a very different behavior. Indeed, no transcriptional output is triggered at low stress levels although DS3-1-02 Hog1p is phosphorylated. At higher stresses the expression becomes bimodal and the fraction of cells activating the Hog1 Activation of mating response converts the HOG pathway specific expression program increases. from a transient to a sustained response Conclusion: We have discovered a novel regulatory mechanism Colman-Lerner, Alejandro1; Baltanás, Rodrigo2 that leads to a bimodal expression output of the signaling 1University of Buenos Aires, Physiology, Molecular and Celular cascade. To identify the molecular mechanism underlying this Biology, Buenos Aires, Argentina; 2CONICET, IFIByNE, Buenos signaling behavior, we are currently comparing the signaling Aires, Argentina response of wild-type and deletion or overexpression of specific pathway components to study the dynamic regulation of the Objective: Cells are usually exposed to multiple environmental signaling cascade and possible implications for bimodality. signals, which often act together to affect cell decisions. Here

32 ICSB 2008 DS3-1-04 Conclusions: These findings demarcate the regions above and below the EC50 on the dose axis and the early and late Regulatory control and the costs and benefits of regions along the time axis as zones of distinct quantitative biochemical noise system behavior. The Vps64 results uncovered the first identified Tanase-Nicola, Sorin1; Ten Wolde, Pieter Rein2 player in a specific mechanism that functions at high pheromone 1University of Michigan, Physics, Ann Arbor, Michigan, United concentrations to reduce noise in signaling. The finding that loss States; 2FOM Institute AMOLF, Amsterdam, Netherlands of Kar3 and other proteins that control the position of the nucleus increases signal variation suggests that nuclear position control Objective: Experiments in recent years have vividly demonstrated may regulate a cell’s response to the pheromone signal. These that gene expression is often highly stochastic. However, how results indicate that the yeast pheromone response is regulated noise in gene expression affects the fitness of an organism is by a number of distinct mechanisms that modulate variability in unknown. Here, we present a model that makes it possible to the transmitted signal. It is possible that some of these regulatory quantify the effect of protein concentration fluctuations on the mechanisms may have evolved in response to selective pressure growth rate of a population of genetically identical cells. for discrimination in mating partner choice. Results: The model predicts that the population’s growth rate depends on how the growth rate of a single cell varies with DS3-1-06 protein concentration, the variance of the protein concentration fluctuations, and the correlation time of these fluctuations. The Fundamental limits on noise suppression 1 2 2

model also predicts that when the average concentration of Paulsson, Johan ; Vinnicombe, Glenn ; Ioannis, Lestas Orals a protein is close to the value that maximizes the growth rate, 1Harvard University, Systems Biology, Boston, United States; Dedicated fluctuations in its concentration always reduce the growth rate. 2University of Cambridge, Engineering, Cambridge, United However, when the average protein concentration deviates Kingdom sufficiently from the optimal level, fluctuations can enhance the growth rate of the population, even when the growth rate of a cell Objective: Negative feedback is common in all types of biological depends linearly on the protein concentration. The model also processes and can increase a system’s stability to internal shows that the ensemble or population average of a quantity, and external perturbations. But at the molecular level, control such as the average protein expression level or its variance, is in loops always rely on finite rates for random births and deaths of general not equal to its time average as obtained from tracing a individual signal molecules and involve delays due to production single cell and its descendants. or transport. We want to mathematically determine how this Conclusions: If the response machinery cannot optimally adjust fundamentally limits noise suppression and experimentally the cell’s protein composition, e.g., when there is a drastic change determine the efficiency of natural feedback loops. in the environment, noise in the protein levels can enhance the Results: By combining mathematical tools from information population’s growth rate. If the response machinery can optimally theory and physical chemistry we show that seemingly mild change the cell’s composition, then there is an evolutionary constraints place severe limits on fluctuations that no type of pressure to minimize the fluctuations in the protein levels. control system could overcome, regardless of nonlinear or However, there is a cost of making the response machinery, and spatial effects. We also show how limits and trade-offs become our analysis predicts that the design of the response machinery dramatically more restrictive in chemical cascades, where is determined by the trade-off between the cost of making the information is inevitably lost at each step. The theory is formulated response machinery, and the benefit of reducing the fluctuations in terms of biological observables that were measured for many in the cell’s protein composition. systems and existing data suggest that cells use brute force when noise suppression is essential, for example expressing regulatory DS3-1-05 genes 10,000s of times per cell cycle. We also developed new experimental methods to count the exact number of plasmid Genetic mechanisms that regulate the precision of molecules per individual cell. The results suggest that plasmids signaling in different regions of dose and time axes of the operate close to the fundamental limits. yeast pheromone response Conclusions: The lower limits on noise are surprisingly high and Pesce, Gustavo1; Yu, Richard1; Rockwell, Daniel1; Colman-Lerner, determined by the quartic rather than square root of the number Alejandro2; Brent, Roger1 of signal molecules made: a decent job is 16 times harder 1The Molecular Sciences Institute, Berkeley, California, than a half-decent job. This may explain many subtle feedback United States; 2Instituto de Fisiología, Biología Molecular y phenomena observed, such as creative strategies to minimize the Neurociencias, UBA-CONICET, Departamento de Fisiología, length of signaling cascades, nested feedback loops that partially Biología Molecular y C, Buenos Aires, Argentina counteract each other, and molecular memory that reduces the randomness of each birth or death event – mechanisms that in Objective: We sought to gain insight into how different inputs turn may explain why the negative feedback loops controlling result in distinguishable outputs at the single cell level. We did plasmid replication are so efficient. so by isolating mutants with altered dynamics of cell-to-cell variation in the response of Saccharomyces cerevisiae to mating DS3-1-07 pheromone. Results: We quantified the output of the pheromone response Control of stochastic gene expression by host factors at in single yeast cells by measuring the ratio between the output of the HIV promoter pheromone inducible and constitutive fluorescent reporter genes Burnett, John1; Miller-Jensen, Kathryn1; Shah, Priya1; Arkin, at different pheromone concentrations (doses). We determined Adam2; Schaffer, David1 that the normalized population average dose response curve 1University of California, Chemical Engineering, Berkeley, United is stable for up to 2 hours. In contrast, cell-to-cell variation States; 2University of California, Bioengineering, Berkeley, United in pathway output (η2(P)) changed over time, displaying two States regimes: at doses lower than the dose that yields half maximal output (EC50), η2(P) quickly stabilized at a relatively high, dose- Objective: After HIV genome integration into the host independent value, but at doses higher than the (EC50), η2(P) chromosome, the viral promoter coordinates a complex set of declined over time to a lower and dose-dependent value. We inputs to control the establishment of viral latency, the onset then devised a high-throughput mutant hunt to identify yeast of viral gene expression, and the ensuing expression levels. genes that affected these behaviors. We found that cells lacking Among these inputs are chromatin structure at the integration the ER tail-anchored protein Vps64 showed the “low-dose” site, host transcription factor levels, and the virally encoded behavior of η2(P) at all doses. Cells lacking the kinesin Kar3 and transcriptional regulator Tat. Importantly, transcriptional noise from other proteins involved in nuclear positioning showed dramatic host and viral regulators may play a critical role in the decision increases in variability. between replication vs. latency, as stochastic fluctuations in gene

ICSB 2008 33 expression are amplified by a Tat positive transcriptional feedback expression for proteins present at low or high copy numbers. loop. Our data provides the first comprehensive proteomic resource of Results: We employ a systems-level approach to dissect the expression and noise with high sensitivity for the model organism roles of cis-regulatory elements within the HIV promoter to gene E. coli. expression dynamics and transcriptional noise. By introducing mutations into numerous cis-acting elements of the integrated DS3-1-09 model HIV viruses, we use flow cytometry, mRNA quantification, and chromatin immunoprecipitation to demonstrate significant Origins of cell-to-cell variability in the life-or-death decision functional differences in contributions of particular sites to viral Spencer, Sabrina; Albeck, John; Burke, John; Sorger, Peter gene expression dynamics. These approaches reveal distinct Harvard, Department of Systems Biology, Boston, United States properties of the two NF-κB sites, with κB site I serving a stronger activating role than κB site II. By regulating the recruitment of Objective: Cells in isogenic populations display significant activating and repressing complexes to the promoter, each of the variability in their fate after stimulation with TNF-family death- three Sp1 sites stabilizes both the active and inactive expression inducing ligands: cells differ in whether or not they die, as well states, such that its mutation promotes noisy gene expression. as when they die. We seek to understand the sources of this Dedicated Corroboratively, analysis of 362 clonal cell populations infected variability, as one of these ligands (TRAIL) is in clinical trials as a

Orals with the viral variants reveals that mutations in any of the Sp1 chemotherapeutic. sites yield a 6-fold higher of frequency of clonal bifurcation Results: To ascertain the origins of variability in the waiting time compared to the wild type promoter. to death, we use live-cell microscopy to perform a cell pedigree Conclusions: Thus, each Sp1 and NF-κB site differentially analysis on timing of death after stimulation with TRAIL. We find contributes to the regulation of viral gene expression, and Sp1 that sister cells show a heritable correlation in death time that sites functionally “dampen” transcriptional noise and thereby decays as a function of time since division, revealing that initial modulate the frequency and persistence of viral latency. This conditions are important for determining the waiting time to systems approach elucidates the synergistic contributions of host death. We therefore examine whether heterogeneity in the basal and viral factors to the dynamics, magnitudes, and stochastic concentrations of proteins in the apoptosis pathway could affect effects in viral gene expression, as well as provides insights into the timing of cell death. mechanisms that contribute to proviral latency. Previous work in the lab has narrowed down the reactions most responsible for death time variability to those upstream of DS3-1-08 the mitochondria. Using live-cell microscopy with fluorescent reporters for initiator caspase activation, we find that variability Single-cell analysis of the Escherichia Coli proteome with in the timing of mitochondrial outer membrane permeabilization single-molecule sensitivity (MOMP) is a result of variability in both the threshold amount Choi, Paul1; Taniguchi, Yuichi1; Chen, Huiyi2; Babu, Mohan3; Emili, of caspase-8 activity required for MOMP and the rate at which Andrew4; Xie, X. Sunney1 caspase-8 activity reaches that threshold. As the threshold for 1Harvard University, Department of Chemistry and Chemical MOMP is governed by Bcl-2 family proteins, we next force a Biology, Cambridge, United States; 2Harvard University, spread in the amount of these proteins and correlate the initial Department of Molecular and Cellular Biology, Cambridge, United level of each with time of MOMP. We find that cells with anti- States; 3Terrence Donnelly Centre for Cellular and Biomedical apoptotic protein levels above a measured threshold amount Research, University of Toronto, Banting and Best Department of are resistant to MOMP. In contrast, modulation of pro-apoptotic Medical Research, Toronto, Canada; 4Terrence Donnelly Centre protein levels reveals that the mean and variance of death time for Cellular and Biomedical Research, University of Toronto, Dept. decrease as protein level increases. of Genetics and Dept. of Med. Research, Toronto, Canada Conclusions: Differences in the probability and timing of cell death can arise from natural variation in the abundance of Objectives: Our goal is to characterize protein expression across regulatory molecules. the entire Escherichia coli proteome using sensitive single-cell imaging. We aim to characterize the system-wide heterogeneity Dedicated session 3-2: Synthetic biology of gene expression across cell populations, which cannot be measured by bulk methods such as microarrays or western blots. Although single-cell flow cytometry has been previously applied DS3-2-01 to other organisms, it lacked the sensitivity to detect many low copy proteins such as transcription factors. Our novel approach Computational engineering of synthetic circuits integrates live-cell single-molecule microscopy with a high- Stelling, Joerg throughput microfluidic platform to reveal properties, functions, ETH Zurich, Zurich, Switzerland and origins of cell-to-cell variation in the E. coli proteome. Results: We constructed chromosomal fluorescent protein Objective: Ultimately, synthetic biology aims at establishing fusions for over 1000 ORFs from the E. coli genome by efficient, novel, useful biological functions by suitably combining well- low-cost conversion of an existing Sequential Peptide Affinity characterized parts. Especially when complex circuits - in [SPA]-tag library. We also developed a microfluidic platform terms of the number of components and interactions involved, for high-throughput fluorescence microscopy, coupled with or with respect to the dynamic behavior - are to be designed, automated image analysis, to record the protein expression of computational engineering methods have to be an integral part of over 100,000 cells per hour, which is sufficient to measure the the approach. statistics of 100 different reporter strains. Results: This talk will focus on engineering concepts to achieve We measure the distribution of protein expression across cell scalability and robustness (relative insensitivity to external or populations and determine noise properties of each gene internal perturbations of the designed circuits). Both are important with single-molecule sensitivity as necessary. We also image concerns for the field because the biology-based parts employed the localization of proteins to the membrane, cytoplasm, and are not (yet) well-characterized, the circuits have to operate in chromosome. To determine possible factors affecting the noise a noisy (cellular) environment, and they cannot be completely of specific genes, we correlate our expression data with other isolated from, e.g., a cellular context. We will discuss tools global data sets. Comparing expression profiles between different for computational design that enable the scalable modeling growth conditions also reveals associations between functional of ‘classical’ synthetic genetic circuits and we will illustrate proteomics and the noise or expression of specific genes. robustness characteristics of engineered circuits with a time-delay Conclusions: We find that a substantial fraction of the proteome circuit in mammalian cells. expresses at low copy numbers and is subject to high noise Conclusions: In perspective, synthetic approaches do not only values. We also observe global properties of protein noise have the potential of major impacts in different application areas, and find differences in the scaling between noise and average but also present challenging problems for engineering design.

34 ICSB 2008 DS3-2-02 Results: We have constructed a library of synthetic genes with randomized codon usage. The synthetic genes all encode the Gene network engineering: Scaffolds for evolution same amino acid sequence (GFP), but contain extensive variation Isalan, Mark1; Garcia de Carvalho, Andreia2; Sanders, Phil2; in their synonymous positions. The library is designed so that it Garriga Canut, Mireia2; Constante, Marco2; Fajardo Sanchez, can be used in any expression system, including in vitro systems Emmanuel2; Herrmann, Frank2; Lemerle, Caroline2; Raineri, as well as bacterial, yeast, and mammalian cells. Applying this Emanuele2; Serrano, Luis2 tool to Escherichia coli, we show that the expression levels varied 1Center for Genomic Regulation (CRG), UPF, EMBL-CRG 250-fold across the library of synonymous genes. The pattern Systems Biology Unit, Barcelona, Spain; 2Centre for Genomic of variation was robust to changes in experimental conditions. Regulation CRG, Barcelona, Spain Nevertheless, traditional measures of codon bias did not explain the extensive variation in gene expression. Predicted RNaseE Synthetic biology aims to understand biological processes cleavage sites explained a small amount of expression variation through reconstruction. In the process, potentially useful (3.5%), whereas the energy associated with mRNA folding devices or systems may be built. We present work on explained nearly half of the expression variation. The influence of engineering a number of biological networks, including artificial folding energy on gene expression was highly local, concentrated transcription factor networks for spatial pattern formation. We around the ribosomal binding site. also demonstrate that tinkering with large-scale networks is Conclusions: Using a synthetic biology approach, we have possible, such as the case of the E. coli transcription network, identified a hierarchy of coding sequence determinants

where ‘rewiring’ the promoter inputs and ORF outputs of major of expression in Escherichia coli. Our results will facilitate Orals hub genes is not only tolerated, but allows the evolution of computational prediction of gene expression levels, and rational Dedicated new properties. By applying similar techniques to those used design of genes with desired expression characteristics. in protein engineering, such as the synthesis of combinatorial libraries around a scaffold and applying selective pressures for DS3-2-05 artificial evolution, one can design and evolve gene networks with particular activities. De novo evolution of promoters in bacteria Hwa, Terence; Li, Sabrina; Liu, Shumo DS3-2-03 UC San Diego, Department of Physics, La Jolla, CA, United States Towards large designed DNA libraries Shabi, Uri Objective: In the post genome era, plasticity in gene regulation is Weizmann IS, Rehovot, Israel recognized as a predominant mechanism generating biodiversity. Results of various comparative genomic analyses suggest that Building faultless objects from faulty components may seem like promoters can evolve rapidly. However, it is often difficult to relate a task for an Alchemist but in fact it is a fundamental challenge the observed changes in promoter sequences directly to the of biology today. Synthetic designed DNA libraries are dearly phenotype and physiology. Here we describe the application of needed in biological and biomedical research. We developed an a procedure with alternating rounds of in vitro mutagenesis and error-correcting recursive construction procedure that addresses in vivo selection to evolve gene expression from a designated this challenges. At the core of our method is a novel general promoter region in E. coli. method of composing two fragments of DNA in a specific way Results: In our study, the selection pressure was gradually and with high yield. To construct long DNA molecules, smaller applied, and the genotype, phenotype and fitness of samples of building blocks are thus composed recursively in a hierarchical the evolving bacteria population were quantitatively characterized. manner. One major issue, as with all DNA synthesis methods, is Starting with a random library of promoter sequences with the high error rate in synthetic oligonucleotides resulting in errors no detectable level of gene expression, significant promoter in every target molecule. Our method naturally deals with these activities were obtained in a few rounds of evolution. The error, by recursively identifying error-free stretches of DNA and emerging promoter sequences typically contained multiple, composing them to reconstruct complete error free molecules. overlapping core promoter motifs. Moreover, a promoter directing We have recently extended our method to construct large libraries transcription in one direction could be easily evolved to transcribe of molecules quickly and efficiently. Our algorithm identifies in the reverse direction with a few point substitutions. Even bi- useful repeating stretches of DNA and maximizes their use as directional promoters could be readily evolved within a region well as identify existing fragments of DNA to further increase the of ~30 nucleotides. An important ingredient in the evolution of efficiency of construction and error-correction. divergent promoters was found to be the minimization of negative To test our method, a library of 100 promoters, with a total length interference due to promoter occlusion. of roughly 40kbps, is being constructed using less than 7kbps of Conclusions: The multiple promoter structure bred by our synthetic DNA. It is designed to explore theoretic mechanisms in evolution process make the system more robust to mutations. the regulation of transcription in Eukaryotes. Similar multi-promoter motifs have been reported for a large To conclude, our method provides a novel and robust foundation number of endogenous promoters in the E. coli genome. for the design and construction of synthetic DNA libraries from The outcomes of these experiments provide unprecedented both synthetic and existing building blocks. Importantly, our demonstrations of the extreme plasticity and evolvability of method is fully amenable to automation which would allow in the bacterial promoters, and suggest possible pathways by which future the construction of massive libraries in parallel and in a more complex promoters may arise. short time. DS3-2-06 DS3-2-04 Molecular titration: A simple mechanism to generate large Coding sequence determinants of gene expression ultrasensitivity Kudla, Grzegorz1; Tollervey, David1; Plotkin, Joshua2 Buchler, Nicolas 1University of Edinburgh, Edinburgh, United Kingdom; 2University Rockefeller University, Center for Studies in Physics and Biology, of Pennsylvania, Philadelphia, United States New York, United States

Objective: The regulation of gene expression is thought to be Objectives: Molecular titration involves the sequestration of an dominated by upstream and downstream non-coding sequences. active protein (A) by protein (B) into an inactive complex (AB). But recent experiments suggest that sequence elements within Intuitively, repressor B serves as a finite sink that can buffer and coding regions may be equally important for regulation. We have titrate low concentrations of A. If sufficient protein A is produced, developed an experimental platform to systematically quantify the then the sink is saturated and A can exhibit a strong threshold effects of synonymous sequence variation on gene expression. or ultrasensitive response. Titration theory demonstrates

ICSB 2008 35 that this ultrasensitivity grows both as a function of repressor DS3-2-08 concentration and increased binding affinity. Results: We took a synthetic approach to test molecular titration How to build a network of synthetic biological oscillators in budding yeast. Using a dimeric, bZIP transcription factor (C/ that synchronize EBP) and dominant-negative proteins that bind C/EBP into an Russo, Giovanni; di Bernardo, Mario inactive complex, we show that molecular titration converts University of Naples Federico II, Systems and computer science, graded transcriptional responses into ‘all-or-none’ responses. Naples, Italy The resulting ultrasensitivity can be equivalent to cooperative processes with large Hill coefficients (nH=8). In addition, we show Objectives: Synchronization is a fundamental mechanism in that molecular titration can lead to transient ultrasensitivity on biological networks: examples include networks of neurons in the short timescales even though the system itself will not exhibit circadian pacemaker and the insulin-secreting cells of pancreas. ultrasensitivity at steady-state. In this paper we examine the problem of finding conditions for Conclusions: Many proteins in natural regulatory networks constructing a network of synthetic biological oscillators that involve the formation of inactive protein-protein complexes, e.g. synchronize. repressor/anti-repressor in bacteriophage, sigma/anti-sigma in Results: Results are obtained using two tools from dynamical Dedicated prokaryotes and bZIP or bHLH transcription factors in eukaryotes. systems theory and linear algebra: Contraction theory [1] and

Orals Our results demonstrate that this simple mechanism can generate Gerschgorin theorem [2]. The key stage in the derivation is the large ultrasensitive responses in vivo, and molecular titration might construction of a so-called virtual system describing the main be playing such a role in these regulatory networks. Moreover, features of the synchronization dynamics of the network under molecular titration in conjunction with feedback can lead to investigation. By appropriate differentiation and analysis of the bistable or oscillatory regulatory networks. We note that our system Jacobian, we establish formal conditions ensuring the approach uses bZIP transcription factors and rationally designed synchronizability of the network. By using Gershgorin theorem, we dominant-negatives. This suggests a strategy for designing derive a set of in parameter space, giving biological guidelines for bistable or oscillatory genetic networks useful for synthetic the construction of synthetic genetic oscillators that synchronize. applications. Conclusions: A new, completely general, algorithmic procedure is presented giving algebraic conditions guaranteeing the DS3-2-07 existence of a common robust synchronous state for an intercellular synthetic biological network. Easy-to-satisfy biological Model-driven synthetic biology guidelines are proposed stemming from a rigorous mathematical Kaznessis, Yiannis1; Sotiropoulos, Vassilios2; Tomshine, derivation. The effects of noise and parameter mismatches Jonathan2; Iyer, Kavita2; Hill, Anthony2 are also investigated. We find that, if for at least one protein or 1University of Minnesota, Chemical Engineering and Materials mRNA of each synthetic circuit of the network, the sum of the Science, Minneapolis, United States; 2University of Minnesota, production/inhibition rate due to other proteins or mRNAs of Minneapolis, United States the circuit is greater than the degradation/dilution rate, then the network will synchronize. Since the production/inhibition rates The nascent field of synthetic biology offers the promise of (i.e. the derivative of the production/inhibition functions) depend designer gene networks that precisely control the expression of on chemical properties of the proteins and mRNAs, it is possible protein molecules. Biomedical and biotechnological applications to use those to inform the design of synchronizable biological abound: from protein production optimization to biosensing and networks. to gene therapies, engineering novel synthetic gene regulatory [1] On Contraction Analysis for Nonlinear systems, W. Lohmiller J. networks is taking advantage of an ever-expanding toolbox of Slotine, Automatica, Vol. 34, pp 683-696 molecular components becoming known thanks to genome [2] Matrix Analysis, R. Horn C. Johnson, Cambridge University projects. Press, 1991, pp 344-350 In the presented work we will describe how to rationalize synthetic biology using model-driven, molecular-level engineering principles. Dedicated session 3-3: Software tools In the presentation we will focus on the theoretical effort to develop an algorithm for simulating biomolecular systems across all relevant time and length scales; from stochastic-discrete to DS3-3-01 stochastic-continuous and deterministic-continuous models, we are developing the theoretical foundation for accurately Challenges for computational systems biology simulating all biomolecular interactions in transcription, translation, Csikasz-Nagy, Attila regulation and induction and how these result in phenotypic CoSBi, Microsoft Research University of Trento CoSBi, Trento, probability distributions at the population level. We are simulating Italy and developing design principles for bio-logical AND gates and tetracycline-inducible networks. The latter are being used in Objectives: We need new type of software tools to deal with gene therapy applications, among other important biomedical new type of biological questions that can be asked after some applications, and we are attempting to address current, known revolutionary developments in experimental research. Future practical limitations. Armed with supercomputers, we are computational software tools should deal with multi scale- predicting the relation between synthetic DNA sequences and multilevel molecular network interactions in space and time. important, biomedically relevant, physiologic behavior. We are Results: Several tools are developed that deal with different parts engineering these networks in E.coli, establishing an integral of this problem. At CoSBi we develop a software tool to treat connection between experimental expression systems and these computational questions piece by piece. Our approaches multiscale models. are initiated from methods developed for computer science and References: based on programming languages. Our novel programming - Y N Kaznessis, Models for synthetic biology, BMC Systems language (BlenX) can easily and intuitively code several biological Biology, 1:47, 2007. processes; especially problems with combinatorial complexity - H. Salis, Y. Kaznessis, “Accurate Hybrid Stochastic Simulation can be easily handled within our conceptual and computational of a System of Coupled Chemical or Biochemical Reactions”, framework. Journal of Chemical Physics, Vol. 122, p. 054103 1-13, 2005. Conclusions: I will describe the state of the art of our tools and - V. Sotiropoulos, Y. Kaznessis, “ Synthetic tetracycline-inducible review some related approaches as well. regulatory networks: computer-aided design of dynamic phenotypes” BMC Systems Biology, 1:7, 2007

36 ICSB 2008 DS3-3-02 analysis which is specific for small changes. Optimization is a strategy to minimize or maximize any feature of the model, i.e. iBioSim: A tool for the analysis and design of genetic its state variables or derived quantities. Together these facilities circuits provide flexible ways to analyse models and their behaviour Myers, Chris1; Barker, Nathan2; Kuwahara, Hiroyuki3; Madsen, in wide range of conditions. Examples of application of these Curtis4; Nguyen, Nam-Phuong4 methods and their combined use will be demonstrated using 1University of Utah, Department of Electrical and Computer models available in the Biomodels database. Engineering, Salt Lake City, UT, United States; 2Southern Utah Conclusions:The use of advanced computational analyses, University, Dept. of Computer Science and Information Systems, and in particular their combined application, provides a powerful Cedar City, UT, United States; 3Microsoft Research - U. of Trento, means to study the behaviour of models in wide ranges of Centre for Computational and Systems Biology, Trento, Italy; parameters. This work demonstrates how this can be achieved 4University of Utah, School of Computing, Salt Lake City, UT, very easily with the software COPASI, i.e. without requirement United States of additional programming. It is hoped that this will enhance the range of model characterization carried out regularly. Objective: iBioSim is a tool that supports learning of genetic circuit models, efficient abstraction-based analysis of these DS3-3-04 models, and the design of synthetic genetic circuits. iBioSim includes project management features and a graphical user Automated image analysis for quantification of protein

interface that facilitates the development and maintenance of localization in budding yeast Orals genetic circuit models as well as both experimental and simulation Kvarnström, Mats1; Logg, Katarina2; Hagmar, Jonas1; Bodvard, Dedicated data records. Kristofer2; Käll, Mikael2 Results: Models in iBioSim can be created using either an 1Fraunhofer-Chalmers Center, Göteborg, Sweden; 2Chalmers SBML editor or a Genetic Circuit Model (GCM) editor. The SBML University of Technology, Applied Physics, Göteborg, Sweden editor and the iBioSim simulation engine support virtually all of SBML Level 2 Version 3 including reactions, rules, events, Objective: Key insights in protein function can be provided by constraints, etc. The GCM editor improves the efficiency of studying the sub-cellular and temporal localization of proteins in model development by supporting modeling at a higher level vivo and fluorescence microscopy provides an excellent tool to of abstraction than the molecular level supported by SBML. conduct such measurements. However, due to the subjective Namely, a GCM includes only important species and their and qualitative nature of human interpretation of images, manual influences upon each other.iBioSim can automatically translate objective and unbiased analysis of images is not straightforward from GCM to SBML models for analysis. A GCM can be either and for large-data sets it becomes practically impossible. manually created or automatically learned from time-series data. Results: We present fast image analysis algorithms for iBioSim also includes an efficient simulation engine that supports automated cell recognition in bright field images of populations ODE, stochastic, and Markov chain analysis of these models. of budding yeast cells. The algorithms do not rely on fluorescent This engine utilizes automatic abstraction to improve analysis staining of the cell membrane and they are therefore in particular time, often by one to two orders of magnitude. Finally, iBioSim suitable for in vivo studies. After the cell contour has been has a graphical editor for visualizing both time series and event determined, assessment of various cell morphology parameters probability analysis results. is straightforward. Spatial analysis of the corresponding Conclusions: iBioSim has been applied successfully to fluorescence image is also possible and includes e.g. numerous projects including an analysis of the phage λ decision determination and classification of protein localizations. We circuit and the E. coli Fim switch. It has also been applied to the exemplify this by identifying protein abundance to three different design of a synthetic genetic Muller C-element, an asynchronous types of spatial configurations corresponding to cell nucleus, state-holding gate. In these and other efforts, the iBioSim tool plasma membrane and peroxisomes. with its support for automatic abstraction has been shown to Conclusions: The cell recognition method is fast and robust greatly improve the productivity of researchers who are analyzing against variations in experimental parameters such as clustering and designing genetic circuits. of cells. It also adapts well to variations in cell density, and illumination level. This makes the algorithms suitable for large DS3-3-03 scale studies where the major part of the analysis has to be done without human intervention. The performance of cell recognition Advanced simulation and analysis of biochemical models and contour extraction was tested on more than 1000 cells in using COPASI 25 images and it was found that 96% of the cells were correctly Gauges, Ralph1; Hoops, Stefan2; Sahle, Sven1; Dada, Joseph3; defined. We also demonstrate how the shape criterion used here Willy, Paul1; Kummer, Ursula1; Mendes, Pedro3 for budding yeast cells can be adapted to other types of cell 1University of Heidelberg, Institute for Zoology/BIOQUANT, shapes such as fission yeast and red blood cells. Heidelberg, Germany; 2Virginia Tech, Virginia Bioinformatics Institute, Blacksburg, United States; 3University of Manchester, DS3-3-05 School of Computer Science, Manchester, United Kingdom Building new experiment-oriented simulation tools Objective: COPASI is a popular systems biology software for Moraru, Ion; Li, Ye; Schaff, James; Cowan, Ann; Loew, Leslie simulation and analysis of biochemical reaction networks. Apart University of Connecticut Health Center, Farmington, United from the well-known basic functions for calculating steady states States and time courses, COPASI is equipped with a number of higher- level analyses that can be used individually or combined to Objective: The aim of this work is to create data- and provide powerful characterizations of models. experiment-driven simulation software that leverages kinetic Results: Here we focus on three advanced tasks of COPASI, models of intracellular processes. namely sensitivity analysis, parameter scanning and sampling, Results: A challenge is the need for generic, interchangeable and optimization. Sensitivity analysis is an method that estimates descriptions of mechanistic hypotheses of cellular physiology how much each parameter of the model affects each variable; involved combined with flexible and detailed descriptions of this allows for an estimation of which parameters may be more experimental protocols and manipulations. We used components important to determine with high accuracy. Parameter scanning and technologies from the Virtual Cell platform (http:/vcell.org/), is a facility in COPASI that facilitates carrying out changes in the which was designed to provide a separation of layers representing values of model parameters and map the resulting state of the biological models, physical mechanisms, geometry, mathematical system. This can be processed in a regular grid or as random models and numerical methods. This separation clarifies the samplings, and it is a way of finding out how the model behaves impact of assumptions and approximations across physiological in the presence of large parameter changes, unlike sensitivity and experimental contexts. We developed standalone software

ICSB 2008 37 packages for kinetic model-assisted simulation, analysis, and for international scientific projects, and to aid scientific journals interpretation of common experimental protocols. One example, in reviewing manuscripts that contain kinetic models. Finally, to focused on quantitative microscopy experiments, is the “Virtual act as a research tool for Systems Biologists who need to make FRAP” tool. It is a mostly data-centric experiment analysis comparative analyses of many kinetic models. framework that allows incorporating fluorescence microscopy Results: Since 2000 JWS Online has provided the Systems (instrumentation, indicators, labels) into existing models of cellular Biology community with a web based simulator for kinetic models physiology. The microscopy data is formally represented, and it of biological systems. The functionality of the simulator includes: is used to derive cellular geometry, distributions of molecules, time course simulation, structural analysis, steady state analysis, and to perform quantitative comparisons with simulated model parameter scans and metabolic control analysis. Results can predictions. This overcomes the limitations of “traditional” be saved in either text or CSV format. The users have access to approaches that use simplified parameterized “models” with the kinetic models in the JWS Online database and also to the closed-form analytical solutions. A different example is the “Virtual models in the Biomodels database (with which we have close Patch Clamp” tool, which is a more protocol- and model-centric collaboration). Recently the JWS Online functionality has been application, allowing incorporation of complex experimental extended with web services, for which requests can be made in manipulations and rigorous physical descriptions of electrical a browser or from an application via HTTP calls. More important Dedicated processes into kinetic models of molecular interactions. This is the possibility to combine web service calls in a workflow as

Orals overcomes the weaknesses of current electrophysiology and can be formulated in Taverna or Mathematica. Currently JWS kinetic modelling software that is limited in handling arbitrarily- Online collaborates with four scientific journals for the reviewing shaped geometries and in coupling electrophysiological and of manuscripts that contain kinetic models: FEBS Journal, IET biochemical processes in the cell. Systems Biology, Microbiology and Metabolomics. JWS Online Conclusions: These customized environments combine powerful is active in data-management and model curation/storage for the experiment description and analysis features with the flexibility of SysMO, and UniCellSys projects. simulating more general classes of spatial models. Conclusions: JWS Online provides an easy to use web-based interface to a growing list of models, (well over 200 models for DS3-3-06 JWS and Biomodels combined). The newly implemented web services make JWS Online an important research tool for Systems Dynamic modeling and multi-experiment fitting with Biologists. JWS Online also provides an important service PottersWheel component via its continuing collaboration with scientific journals Maiwald, Thomas; Timmer, Jens and research initiatives. Freiburg Center for Data Analysis and Modeling, Physics Department, Univ. of Freiburg, Freiburg, Germany DS3-3-08

Objective: PottersWheel has been developed to provide an Software tools — a survey from a theoretical perspective intuitive and yet powerful framework for dynamic data-based Jirstrand, Mats modeling in Systems Biology. Fraunhofer-Chalmers Centre, Gothenburg, Sweden Results: The key functionality of PottersWheel is multi- experiment fitting, where several experimental data sets are fitted Objective: This talk is an attempt to give a brief introduction to simultaneously in order to improve the estimation of unknown the field of software tools for systems biology with a bias towards model parameters, to check the validity of a given model, and the tools presented in this session, which will mainly be done with to discriminate competing model hypotheses. The program is a theoretical perspective. Furthermore, the abilities and features designed as a MATLAB toolbox and includes numerous user of two specific software tools will be presented; PathwayLab interfaces. For example using sliders, parameter values or — a software for component based modeling and simulation characteristics of input functions like extracellular stimuli, can of biochemical reaction networks and NLMEtools — a Matlab be changed continuously in real time. Dynamically generated C toolbox for nonlinear mixed effects modeling using stochastic MEX files of the differential equations and the use of FORTRAN differential equations. integrators provide fast simulation and fitting procedures. A Results: PathwayLab facilitates the modeling process by a comprehensive application programming interface is available for simple to use drag-and-drop graphical user interface where customization and use within own MATLAB scripts. reaction networks quickly are assembled using readymade or Detailed analysis of parameter-identifiability using the MOTA user defined modeling objects. We give an overview of existing approach [1] detects structural limits of parameter estimation. modeling objects, their mathematical meaning, and the current Stastistical tests comparing rival models can be used to rule out simulation capabilities of the tool. The complete set of equations wrong models and to exemplify what can be concluded from and parameters of a model can be exported for use in more experimental data given a certain noise level. External driving general computational platforms such as Mathematica and input functions help to optimize the expected information of Matlab. In particular, the connectivity to Mathematica and its further experiments. dynamic interactivity opens up interesting possibilities, which will Conclusions: PottersWheel is freely available for academic briefly be discussed. usage from www.PottersWheel.de. It is intensively used by In the case that data from biological experiments come from a experimentalists and modelers since 2005. population containing several individuals, such as population [1] S. Hengl, C. Kreutz, J. Timmer, and T. Maiwald, Data- imaging data of yeast cells or multi-experiment test animal set Based Identifiability Analysis of Nonlinear Dynamical Models. ups, there are both inter and intra individual variations. A statistical Bioinformatics, 2007, 23: 2612-8 framework that takes this into account is so called nonlinear mixed effects modeling. The NLMEtools toolbox provides means DS3-3-07 for model specification, simulation, and parameter estimation utilizing this kind of data. We will give a brief overview of its JWS Online: A web-accessible model database, simulator features and show a simple example illustrating its use. and research tool Conclusions: The software tools survey will be concluded with Du Preez, Franco1; Snoep, Jacky1; van Gend, Carel1; Conradie, an attempt to point at directions for software development, which Riaan1; Penkler, Gerald1; Stoof, Cor2 is thought to be of great value for the system biology community. 1University of Stellenbosch, Biochemistry, Stellenbosch, South Africa; 2Vrije Universiteit, Cellular BioInformatics, Amsterdam, Netherlands

Objectives: Firstly, to create an easy to use simulation interface to a database of kinetic models, and to act as a central repository for such models. Secondly, to act as a service tool

38 ICSB 2008 Dedicated session 3-4: Model driven well to the dataset. We illustrate this problem using a previously experimental planning published model of apoptosis signalling. Inference based on standard experiments, i.e. perturbing genes one-by-one, is shown to yield networks with the wrong structure although its predictive DS3-4-01 ability is validated using independent validation data. We present an iterative algorithm for experiment design that guarantees Mouse gene function prediction and complex disease sufficient excitation of all network modes and demonstrate its Tasan, Murat1; Tian, Weidong1; Oestergaard, Mikkel2; Tyrer, effectiveness. Jonathan2; Morrison, Jonathan2; Hill, David3; Blake, Judy3; Conclusions: Systematic design of perturbation experiments, Ponder, Bruce2; Easton, Douglas2; Pharoah, Paul2; Roth, where several genes are perturbed simultaneously in a controlled Frederick1 fashion, is necessary in order to infer the true structure of GRN 1Harvard Medical School, Biol Chem Molec Pharm, Boston, MA, from expression data. It is likely that many inferred network United States; 2Strangeways Research Laboratories, Cambridge, models with validated predictive properties have falsely identified United Kingdom; 3The Jackson Laboratory, Bar Harbor, ME, gene interactions. United States DS3-4-03 Objective: Several years after sequencing of the human and mouse genomes, functions for human and mouse genes Quantitative analysis of the mitochondrial-pathway

remain largely undiscovered. A major challenge is to focus requirement for receptor-mediated apoptosis in single cells Orals limited experimental resources on the most likely hypotheses Aldridge, Bree1; Gaudet, Suzanne2; Albeck, John2; Sorger, Peter2; Dedicated using computational predictions of gene function and functional Lauffenburger, Douglas3 relationships. 1Massachusetts Institute of Technology, Biological Engineering, Results: We predicted Gene Ontology (GO) and phenotype Boston, MA, United States; 2Harvard Medical School, Department terms for 21,603 Mus musculus genes, using a diverse collection of Systems Biology, Boston, MA, United States; 3Massachusetts of integrated data sources (including expression, interaction, and Institute of Technology, Department of Biological Engineering, sequence-based data), and an optimized combination of ‘guilt- Cambridge, MA, United States by-profiling’ and ‘guilt-by-association’ inference methodologies. Predictions were evaluated using a held-out gene set, and top Objective: It is generally thought that cells undergo apoptosis predictions were examined manually using available literature. using one of two primary pathways and are classified accordingly A set of ‘functionally linked’ SNP pairs was generated from as being either Type I (mitochondria-independent) or Type II prediction scores and used to focus a search for human SNP (mitochondria-dependent). The control point is thought to be pairs associated with breast cancer status. In a sample of at the level of initiator caspase activation, and Type II cells are 4,000 breast cancer cases and 4,000 controls with ~13,500 identified as those protected from death-ligand induced apoptosis SNP pairs tested, the false discovery rates were 83% among by Bcl2 overexpression. To investigate the conditions that classify non-functionally linked SNP pairs but reduced to 50% among cells as being Type I or II, we used a systems approach that functionally linked SNP pairs (at a nominal p-value of 2E-4). combines kinetic modeling and detailed dynamic analysis with Conclusions: We assigned a confidence score to each gene/ quantitative data of the concentrations and kinetics of proteins term combination. Nearly every GO term achieved greater than involved in this cell-decision network. 40% precision at 1% recall. Among the 36 novel predictions for Results: We used a Lyapunov exponent–based analysis of a GO terms and 40 for phenotypes that were studied manually, physicochemical model of receptor-mediated apoptosis to define >80% and >40%, respectively, were identified as accurate. a multi-dimensional separatrix that demarks Type I and Type II A combination of ‘guilt-by-profiling’ and ‘guilt-by-association’ cell states. Informed by our analysis, we hypothesized that this outperforms either approach alone. We have shown that control may also occur at the level of effector caspases, and functional linkage graphs can be used to enrich for SNP pairs that incomplete activation of the caspases can be observed in exhibiting complex association with breast cancer susceptibility. cells that are neither Type I nor Type II. We perturbed the ratios of caspase-3 to XIAP and measured the kinetics of caspase-3 DS3-4-02 activity with flow cytometry and FRET-based reporters in live-cell microscopy. We experimentally observed the breakdown of the Inference of interampatte gene regulatory networks - with characteristic “all-or-nothing” activation of the caspases in Bcl-2 application to apoptosis signalling overexpressing cells partially depleted of XIAP. Further, we show Nordling, Torbjörn E.M.; Jacobsen, Elling W. that Type II cells can be converted to Type I cells by depletion of KTH - Royal Institute of Technology, Automatic Control, XIAP. Stockholm, Sweden Conclusions: These analyses and data support our hypothesis that the control of Type I/II pathway usage can be controlled at Objective: Inference of gene regulatory networks (GRN) from the level of effector caspases. Furthermore, our results define quantitative expression data has the potential to reveal all a continuum of pathway-usage phenotypes that delineate interactions existing within a selected set of genes. However, conditions leading to partial caspase activation which may microarray data typically only contain a few characteristic modes lead to genomic instability and potential for oncogenesis. This or eigengenes, even when a large number of arrays are recorded characterization may provide a framework to evaluate best use- at varying experimental conditions. The reason and implications cases of IAP-, TRAIL-, and Bcl-2-targeted cancer therapeutics. of this inherent rank deficiency has largely been neglected, even though rank deficiency caused by fewer experiments than measured genes has been addressed. We explain why the data in the former case are rank deficient, what it implies for network inference, and how to counteract it through experiment design. Results: We define interampatte systems as systems characterised by strong interactions necessary to both amplify and attenuate different signals at multiple time-scales. GRN are interampatte with strong directional dependence. This generic network property make microarray data rank deficient and gives rise to features observed as characteristic modes, eigengenes and co-expressed genes. While few modes imply that low order models can be used for data compression and prediction, it effectively prevents inference of causal interactions, since many sparse networks with completely different structure fit equally

ICSB 2008 39 DS3-4-04 DS3-4-06

Model-based study of life and death pathways diverging Fathead minnow steroidogenesis- in vitro modeling and from the CD95 Death Receptor experimentation reveals global regulation of sex hormone Eils, Roland1; Neumann, Leo2; Beaudouin, Joel2; Golks, synthesis Alexander3; Krammer, Peter H.3; Lavrik, Inna N.3; Pforr, Carina3 Shoemaker, Jason E.1; Reyero Vinas, Natàlia G.2; Gayen, Kalyan1; 1Institute for Pharmacy and Molecular Biotechnology, University Perkins, Edward3; Doyle III, Francis J1 of Heidelberg, Department for Bioinformatics and Functional 1University of California, Santa Barbara, Dept of Chemical Genom, Heidelberg, Germany; 2German Cancer Research Center, Engineering, Santa Barbara, CA, United States; 2University of Division of Theoretical Bioinformatics, Heidelberg, Germany; Florida, Department of Physiological Sciences, Gainesville, FL, 3German Cancer Research Center, Division of Immunogenetics, United States; 3US Army Engineer Research and Development Heidelberg, Germany Center, Vicksburg, MS, United States

It is an unresolved dilemma in cellular signaling that triggering Objective: In 2007, the National Research Council called for a of the death receptor CD95 (Fas/APO-1) in some situations shift in traditional toxicology from a strictly observational study Dedicated results in cell death and in others leads to the activation of NF-κB of disease-specific modelsin vivo, to a more predictive science Orals resulting in cell proliferation. We established an integrated kinetic focusing on mechanism-based, biological observations in vitro mathematical model of CD95-mediated life and death signaling. using high throughput technologies. In a collaborative effort Systematic model reduction resulted in a surprisingly simple between the EPA, UCSB, UF, and the US Army, we developed a model well approximating experimentally observed dynamics. mechanistic understanding of the ovarian steroidogenic response The model postulates a novel link between c-FLIPL cleavage in to endocrine disrupting compounds (ketoconazole and fadrozole) the Death-Inducing Signaling Complex (DISC) and the NF-κB of the fathead minnow (Pimephales propelas). Steroidogenesis, pathway. We validated experimentally that CD95 stimulation production of sex hormones from cholesterol, in the fathead resulted in binding of p43-FLIP to the IKK complex followed by its minnow provides an ecologically relevant model to understanding activation. Furthermore, we showed that the apoptotic and NF-κB robustness, compensation and failure of highly dynamic networks. pathways diverge already at the DISC. Model and experimental Results: Steroid production is a highly dynamic process, analysis of DISC formation showed that a subtle balance of consisting of several layers of regulation at the genomic, c-FLIPL and procaspase-8 determines life/death decisions in a proteomic and inter-organ levels along the hypothalamus- non-linear way. pituitary-gonadal axis. Integrating transcriptomics, metabolomics, and previous mathematical models, a hybrid model using ODE’s DS3-4-05 to account for inter-organ signaling and flux balance to describe metabolite dynamics was developed over several iterations in Optimizing 13C labeling experiments for flux elucidation which each iterate of experimental design was influenced by Schellenberger, Jan1; Choi, Wing2; Palsson, Bernhard2 different model characteristics (signaling timing, steady-state 1UC San Diego, Bioinformatics, La Jolla, CA, United States; 2UC behaviors, etc.). The results reveal several feedback mechanisms San Diego, Bioengineering, La Jolla, CA, United States within the ovary that optimize the processing of steroid precursors. Furthermore, though the model is descriptive of the Objective: Carbon-13 labeling experiments have been widely in vitro system, the model reveals, and in vivo experimentation used to study flux rates through metabolic networks. Because confirms, that compensation machinery ensuring robust steroid of high material and time costs in running these experiments, production is not localized to the ovary. it is beneficial to optimize the experimental parameters Conclusions Results show that while local machinery within the beforehand. We present an algorithm which incorporates the ovary seeks to optimize precursor processing, inter-organ global current knowledge about a metabolic network and returns the regulation occurs to compensate for local inhibition. We conclude optimal design parameters to meet an experimental objective. that the fathead minnow has evolved with a distributed strategy This procedure is general enough to answer a wide variety of to protect against acute network failure, and that the application questions about the network such as: The absolute flux through a of a robustness cost function for model discrimination can be particular reaction, the ratio of fluxes between two reactions (flux misleading without proper experimental guidance. Finally, results splits), or how to globally maximizing the confidence in reaction support using multi-organ/compartment models to examine local rates. and global phenomenon. Results: Using a constraint based analysis approach, a set of candidate flux distributions is chosen by Monte Carlo sampling. DS3-4-07 For each flux distribution, a13 C experiment is simulated, giving a distribution of predicted measurements. A score is assigned to Parameter estimation for a circadian clock model in this experiment based on how well the experimental results are neurospora using optimal experiment design able to predict the flux distributions which they came from. This Keienburg, Jens1; Busch, Hauke2; Hoffmann, Christian3; Di procedure is repeated for a variety of 13C labels to maximize the Ventura, Barbara1; Schloeder, Johannes3; Brunner, Michael4; experimental score. Bock, Hans-Georg3; Eils, Roland2 All combinations of single and double labeled 13C glucose were 1University of Heidelberg, Bioquant, Heidelberg, Germany; tested in this fashion on a large-scale E. coli isotopomer model 2German Cancer Research Center (DKFZ), Theoretical with a variety of experimental objectives (for example determining Bioinformatics, Heidelberg, Germany; 3University of Heidelberg, the rate of specific reactions in the network). The experimental IWR, Heidelberg, Germany; 4University of Heidelberg, BZH, score between the best and worst glucose label was highly Heidelberg, Germany significant indicating that choice of label has a strong effect on the results of the experiment. No globally optimal 13C label was found Objective: Quantitative understanding of dynamic biological meaning that the best label is dependent on the experimental systems often involves a model reconstruction based on ordinary objective. differential equations (ODEs). Since their parameter space is Conclusion: Due to the complexity of isotopomer models, largely under-defined and kineticin vitro or in vivo studies are designing the best experiments to answer specific questions time-consuming and expensive, it is important for the estimation is not trivial. Our algorithmic framework is able to quantitatively of kinetic parameters to optimize experiments towards a evaluate different experimental parameters to choose the one maximal yield of novel information. Given a biological system which will be most informative. optimal experiment design (OED) calculates optimal time points and optimal external controls telling the experimentalist how to perform measurements with respect to time, system components, experimental conditions and stimuli. We use the core of the circadian clock in Neurospora - i.e. the oscillations of frq-mRNA,

40 ICSB 2008 FRQ- and WCC-proteins - to show that usage of OED leads to parameter estimates of high confidence with less experimental effort, as compared to intuitively planned experiments. Results: The Goodwin model, which describes the core of the circadian clock with only three ODEs, is combined with an Arrehnius-type temperature dependence of the kinetic parameters to account for the temperature sensitive mechanism in clock adaptation. Based on an initial model with literature-derived kinetic parameters, OED suggests time points for frq-mRNA and FRQ-protein abundance measurements, using the temperature as control parameter to externally influence the expression of the circadian clock gene frq. The returned parameter set is then used to iteratively design a new experiment yielding measurements of higher kinetic information content. After several experimental cycles the A-criterium (average standard deviations of the parameters) is significantly higher compared to intuitively conducted experiments. Conclusions: Application of OED in the core oscillation of the

Neurospora circadian clock demonstrates in a proof of principle Orals experiment its high potential for systems biology: obtaining more Dedicated reliable parameter estimates with less effort. The relevance of this technique will be ultimately shown, when applied to large and close to life in vitro and in vivo models.

DS3-4-08

Model driven experiment planning for the Robot Scientist Clare, Amanda1; Aubrey, Wayne2; Byrne, Emma1; Liakata, Maria1; Markham, Magdalena2; Rowland, Jem1; Soldatova, Larisa1; Sparkes, Andrew1; Whelan, Ken1; Young, Mike2; King, Ross D.1 1Aberystwyth University, Dept of Computer Science, Aberystwyth, United Kingdom; 2Aberystwyth University, IBERS, Aberystwyth, United Kingdom

Objective: The Robot Scientist is a project that aims to automate the process of science. The system automatically generates hypotheses, designs experiments that will test the hypotheses, physically executes these experiments on laboratory automation equipment, analyses the results and uses the new knowledge to generate the next round of hypotheses. The experimental domain for our first Robot Scientist is yeast functional genomics. The generation of hypotheses and experiments is driven by a logical model of yeast metabolism. Results: The Robot Scientist has been physically implemented as a large laboratory automation system. Abduction and conventional sequence similarity bioinformatics have been used to automatically infer hypotheses from the models of yeast metabolism. Deduction from the models was then used to automatically design experiments using yeast gene-deletion strains. These experiments have been laid out in Latin square arrangements and executed by the robotic equipment. Statistics and machine learning methods were applied to determine the results of the experiments and the validity of the inferred hypotheses. Conclusions: The Robot Scientist has rediscovered existing knowledge about enzymes in the aromatic amino acid pathway, and has discovered new knowledge about other enzymes whose parent genes are currently unknown, including several in the lysine biosynthesis pathway. The data describing the model, hypotheses, experiments, results and conclusions is available online and is formalised for further processing.

ICSB 2008 41

Arenas Arenas Arenas A-03

A-01 RiboSys - systems biology of RNA metabolism in yeast Alexander, Ross1; Barrass, David1; Bertrand, Edouard2; Beggs, SYSBIOMED - Systems biology for medical applications Jean3 Schuerrle, Karsten 1University of Edinburgh, Centre for Systems Biology and Society for Chemical Engineering and Biotechnology, Frankfurt, Wellcome Trust Cen, Edinburgh, United Kingdom; 2CNRS, Germany Montpellier, France; 3University of Edinburgh, Centre for Systems Biology and Wellcome Trust Cen, Edinburgh, United Kingdom Objective: SYSBIOMED is a EU-funded Strategic Support Action within FP6. Its core objective is to explore the potential of systems ‘RiboSys’ is an EC-funded project that aims to use system biology for medical research, therapy and drug development. The biology approaches to model both pre-messenger RNA and pre- other main goal is the formation of a network of young scientists ribosomal RNA metabolism in Saccharomyces cerevisiae. Using who define the framework for future research programmes in a series of reporter genes with features that facilitate quantitative ‘Medical Systems Biology’ (MSB). analysis and some which incorporate mutations that affect pre- Preparatory and disease focused exploratory workshops form mRNA processing we are able to quantify RNA precursors and the core of SYSBIOMED. They offer valuable opportunities for determine their rates of production, processing and degradation young scientists to enter this field, for theoreticians to meet through the various post-transcriptional pathways. Through a experimentalists, for industry researchers to meet academic combination of QRT-PCR, in vitro transcribed reporter RNAs, scientists and should help to bridge the gap between the scientific Arabidopsis thaliana spike mRNA and DNA extraction we are communities in systems biology and medical/clinical research. able to estimate the efficiency of RNA extraction, RT-PCR and cell

Arenas Results: SYSBIOMED is providing valuable strategic information lysis, allowing quantification of the reporter transcripts in copies for the assessment of MSB’s potential, thus laying groundwork for per cell. future research initiatives. More information and workshop reports Starting with ab-initio models our modelling partners produced are available at http://www.sysbiomed.org mathematical representations of the processing and degradation of pre-mRNAs. Further refinement of the model parameters has A-02 been performed using the quantitative experimental data. By manipulating the model parameters, predictions will be VALAPODYN: A new systems biology approach to develop made about the behaviour of the systems and we will test these predictive dynamic models of complex intracellular experimentally using yeast mutants that block specific steps e.g. networks for neurological disease spliceosome assembly, first and second steps of pre-mRNA Sanoudou, Despina1; Depaulis, Antoine2; Vafiadaki, Elizabeth3; splicing and RNA turnover. Individual transcripts have also Jackers, Pascale4; Kel-Margoulis, Olga5; Wingender, E5; been visualised by a RiboSys partner, E. Bertrand, to determine Vujasinovic, Todor6; Greenberg, David7; Soreq, Hermiona7; whether the population data reflect the situation in individual DePaw, Edwin4; Dumas Milne Edwards, Jean-Baptiste6 cells. Utilising such quantitative approaches will allow a better 1Biomedical Research Foundation Academy of Athens, Molecular understanding of the relationships between different steps, Biology, Athens, Greece; 2INSERM U836, Grenoble, France; activities and factors in the RNA metabolic pathways than has 3Biomedical Research Foundation Academy of Athens, Athens, previously been achieved by qualitative analyses and intuitive Greece; 4University of Liege, Liege, Belgium; 5Biobase-Gmbh, interpretations. Wolfenbuettel, Germany; 6Helios, Paris, France; 7Hebrew University of Jerusalem, Jerusalem, Israel A-04

Objective: VALAPODYN, a European Commission funded The Manchester centre for integrative systems biology research network, is using an original systems biology approach (MCISB) for the development of an innovative dynamic model of molecular Weichart, Dieter1; Swainston, Neil2; Smallbone, Kieran2; interaction networks (MIN) in relation to cell death and survival for Simeonidis, Evangelos1; Jameson, Daniel1; Dunn, Warwick2; the detection of new therapeutic targets for human neurological Carroll, Kathleen1; Spasic, Irena1; Malys, Naglis1; Khan, Farid1; diseases. To this end, a comprehensive multidisciplinary strategy Broomhead, David1; Gaskell, Simon1; McCarthy, John1; Mendes, has been established combining functional genomics, proteomics Pedro1; Oliver, Stephen3; Paton, Norman1; Westerhoff, Hans1; and bioinformatics. Kell, Douglas1 Results: Using a mouse model of induced hippocampal sclerosis 1MCISB, MIB, University of Manchester, Manchester, United associated with focal epilepsy, dynamic expression analyses Kingdom; 2MCISB, MIB, University of Manchester, Manchester, are conducted at different time points. Proteomic databases United Kingdom; 3University of Cambridge, Department of are being used along with advanced microarray and proteomics Biochemistry, Cambridge, United Kingdom platform systems to investigate protein-protein interactions and regulation networks, identify and validate biological targets The Manchester Centre for Integrative Systems Biology (www. in complex intracellular pathways. The first phase involves mcisb.org, MCISB) at the University of Manchester aims to whole genome and proteome analysis, integrating biological pioneer the development of new experimental and computational and statistical data in order to functionally annotate genes and technologies in Systems Biology, and to aid in their exploitation. proteins. Using Affymetrix microarrays, 2D-DIGE and MALDI/TOF- The MCISB is intended to provide a hub for cutting-edge systems TOF, we are evaluating whole genome and proteome expression biology research in the Manchester area and beyond, acting profiles bringing to light critical new pathways and molecular as a focal point for the creation of the necessary ideas and targets implicated in neurodegeneration. infrastructure, and for establishing new methods and routines. Conclusion: VALAPODYN develops a dynamic and quantitative The closely associated Doctoral Training Centre provides training analysis method for new therapeutic targets through MIN for the new generation of Systems Biologists (see www.mcisb. dynamic models and specifically addresses the systems biology org/dtc) of complex cellular pathways and transcriptional networks. In the research carried out at the MCISB, the bakers’ yeast Novel predictive dynamic models will be validated by testing the (Saccharomyces cerevisiae) is being used as a model selected drug targets on innovative in vivo and in vitro models of organism, because this organism is highly amenable to genetic CNS pathologies. VALAPODYN will provide a cutting-edge highly manipulations and to a variety of high-throughput and genome- accurate in silico tool for identifying novel and effective therapeutic scale omics technologies, thus offering an excellent starting targets in a faster, more efficient and more economical way than it position for demonstrating the principles and methods of Systems is possible today. Biology. The technologies developed and demonstrated in yeast will then be applied to other organisms, including humans. One of the initial challenges is to mine existing sources of

44 ICSB 2008 information and to integrate these with the necessary data groups of industrial microorganisms: yeast (Saccharomyces models. The second challenge is to generate experimental cerevisiae), filamentous fungi Aspergillus( niger and Penicillium protein and metabolite data and the kinetic parameters of chrysogenum) and lactic acid bacteria. The research activities metabolic enzymes. These data will then, together with the data in the Kluyver Centre 2008-2012 are organized in five focused, from existing sources, be introduced into computer models complementary and interlinked research topics within these representing all the metabolic reactions known in yeast. This programmes and have been carefully chosen to efficiently develop integration represents a third type of challenge, and will lead and utilize genomics technology for solving scientific problems to computer models of parts of living cells with predictive and related to the Dutch fermentation industry. explanatory powers. Programme 1: Yeast for chemicals, fuels and beverages The “arena” presentation of the MCISB will showcase the Programme 2: Filamentous fungi for proteins and peptides approaches utilized and the kinds of results that have been Programme 3: Lactic acid bacteria for fermented foods and obtained to date. The centre is truly interdisciplinary in its food ingredients structure, and combines biochemical experimentation, principled Programme 4: Systems Biology of industrial micro- data management and computational modelling to allow a organisms detailed understanding of large and complex biological systems. Programme 5: Industrial Genomics for Society. In the Systems Biology programme, the fundamental principles A-05 that underlie the strategies that microorganisms adopt during long-term adaptation to specific conditions will be addressed. APO-SYS consortium apoptosis systems biology applied to Long-term cultivation will lead to optimal behavior from the Cancer and AIDS: An integrated approach of experimental perspective of the cells, which will not necessarily (and in many biology, data mining, mathematical modeling, biostatistics, cases only by accident) coincide with industrial demand. Using systems engineering and molecular medicine Systems Biology approaches, the aim of this programme is to Imreh, Gabriela better understand the trade-offs that cells face when adapting to APO-SYS Consortium, www.apo-sys.eu, Stockholm, Sweden either constant or dynamic conditions. The model organisms of

this programme are S. cerevisae and L. lactis. In S. cerevisiae, Arenas The APO-SYS Consortium is made up of 23 partners from 10 the short and long term response to temperature variation will European and 2 associated countries (Israel and Switzerland) with be studied. This work will be carried out by the Delft University of the aim of uniting a critical mass of investigators specialized in Technology and the Free University of Amsterdam; in L. lactis the biology, biomedicine, biomathematics and biostatistics, a critical impact of different glucose concentrations will be studied within mass that might solve important, disease-relevant problems in the Top Institute Food and Nutrition and the Free University of the systems biology of apoptosis, placing special emphasis on Amsterdam. cancer and AIDS. The APO-SYS Consortium aims at obtaining major progress in A-07 the comprehension of apoptosis (and more general cell death) in human disease, by combining a series of systems biology Cytoscape: Open source software for network informatics approaches, in silico, in vitro (in organello and in cellula), in vivo Workman, Chris1; Molenaar, Piet2; Adler, Annette3; Kuchinsky, and by integrating experimental results with large data sets Allan3 acquired on tissue samples from patients suffering from diseases 1Technical University of Danmark, Center for Biological Sequence that are caused by deregulated apoptosis, in particular cancer Analysis, Lyngby, Denmark; 2Academic Medical Center, University and AIDS. APO-SYS Consortium will specifically focus on of Amsterdam, Human Genetics, Amsterdam, Netherlands; developing a relevant model repository consisting of known as 3Agilent Technologies, Inc, Santa Clara, California, United States well as newly identified pathways and gene regulatory networks associated with apoptosis, and on the role of specific changes in Analysis of molecular networks has exploded in recent years key genes/gene products in these pathways. due to the wide variety of new technologies that have been The unique combination of mathematical, biomedical and introduced into modern biology. This number of new molecular biostatistical knowledge present in this consortium will lead to the interaction measurements requires a bioinformatics framework generation of large data banks as well as new methods of data to filter, integrate, and interpret the resulting diverse data in a analyses with the final result of refining the molecular diagnosis biologically meaningful way. One emerging and increasingly of major diseases, of optimizing the calculation of prognostic standard approach to doing so is network informatics: integrating and predictive parameters, and of guiding new strategies for the these diverse data into networks for further analysis. A major amelioration of existing treatments and the identification of novel goal of network informatics is to organize and place molecular targets for therapeutic modulation of apoptosis. interactions into models of signaling pathways, protein The APO-SYS Consortium was funded by the European Union complexes, cell structural components, and other cellular under the 7th Framework Programme (FP7/2007-2013) under machinery. The pursuit of this goal is the primary aim of the agreement no. HEALTH-F4-2007-200767. Cytoscape1 open source software project (www.cytoscape.org). Cytoscape is a publicly-available bioinformatics resource for A-06 integration, visualization and query of biological networks to derive computational models. Since late 2001 Cytoscape has Kluyver centre for genomics of industrial fermentation: grown to become a standard tool for biological network analysis. Systems biology programme Its success owes mainly to its timeliness (one of the first tools for Teusink, Bas1; Hugenholtz, Jeroen1; Molenaar, Douwe1; Bakker, visualization of protein networks), open development model (one Barbara2; Westerhoff, Hans2; de Winde, Han3; Pronk, Jack3; of the only such tools that is open-source), and simple plug-in Heijnen, Sef3 interface (that attracted many third-party developers and industrial 1Kluyver Centre for Genomics of Industrial Fermentation, TI Food partners). and Nutrition, Wageningen, Netherlands; 2Kluyver Centre for Cytoscape usage has increased at a rate of approximately 50% Genomics of Industrial Fermentation, Vrije Universiteit Amsterdam, per year and was downloaded >2,000 times per month during Amsterdam, Netherlands; 3Kluyver Centre for Genomics of the first half of 2008 and a total of nearly 55,000 since 2004. It Industrial Fermentation, Delft University of Technology, Delft, has been featured in several hundred recent publications and Netherlands several international databases have interfaced to Cytoscape for visualization and analysis of interactions, including Reactome, The mission of the Kluyver Centre for Genomics of Industrial MiMI, BIND, and IntAct. It is supported by an active discussion list Fermentation (www.kluyvercentre.nl) is to enable the required of Cytoscape users ([email protected]) . breakthrough innovations in industrial biotechnology by The Cytoscape Arena will highlight Cytoscape in use, through a harnessing existing and novel Genomics techniques for high- set of illustrative case examples that show Cytoscape used in the quality, pre-competitive and focused research on three key actual flow of doing biology. Focus will be on the added value of a

ICSB 2008 45 network representation and analyses. e.g. metabolites, genes, proteins and facts e.g. protein-protein interactions, etc., thus successfully linking text with biological 1. Cline, M. et al. Integration of Biological Networks and Gene knowledge. NaCTeM has developed a range of TM services Expression Data using Cytoscape. Nature Protocols 2(10) 2366- and tools which are described and accessed from the NaCTeM 82 (2007). website at www.nactem.ac.uk. NaCTeM’s own tools are offered to users as Web Services. Among the TM tools and services offered A-08 by NaCTeM and of interest to systems biologists are: TerMine, AcroMine, KLEIO, MEDIE, InfoPubMed and FACTA. The NaCTeM CCO: Cell cycle ontology, an integrated knowledge base to Arena will host demonstrations of these tools and discussion of study cell cycle control how they can be used to assist in day-to-day systems biology Kuiper, Martin1; Antezana, Erick2; Mironov, Vladimir3; Blonde, research. Visitors will also have the chance to meet members of Ward2; Egana, Mikel4; Stevens, Robert5; De Baets, Bernard6 our team to discuss individual challenges and the potential for text 1NTNU, Biology, Trondheim, Norway; 2VIB, Gent, Belgium; mining to provide solutions. 3NTNU, Trondheim, Norway; 4Manchester University, Manchester, United Kingdom; 5Manchester University, Manchester, United A-10 Kingdom; 6Gent University, Gent, Belgium CoSBi the microsoft research - university of trento, centre Objective: The DIAMONDS project aimed to establish systems for computational and systems biology biology to analyze cell cycle control, focused to assess the CoSBi dynamics of cell cycle control. This has been done through Microsoft Research University of Trento, Centre for Computational analysis of protein complex assembly and transcriptomics, Systems Biology, Trento, Italy

Arenas mathematical modeling, software development and the construction of an integrated cell cycle application ontology The Centre is a non-profit limited liability consortium shared (CCO). One of the aims of the project was to highlight the 50% by the University of Trento and 50% by Microsoft Research importance of such semantic approaches to study the cell cycle, Cambridge. CoSBi’s main goal is to perform research activities and illustrate this with examples of querying and automated and develop specific languages and mechanisms of modeling, reasoning. analyses and simulations in the fields of medical science, biology, Results: A major accomplishment of the DIAMONDS project bio-information and of complex systems in general. The Centre was the development of a data and ontology integration pipeline principally focuses on research in the field of innovative computer for cell cycle control: in its first phase some existing ontologies science technology as well as on computational models of were integrated. Then, protein and gene data was added. Finally, biological regulatory networks. The researchers at CoSBi are a semantic improvement task is undertaken. The resulting Cell selected through an international recruitment process. At the Cycle Ontology (CCO) was designed to provide a richer view moment there are 23 researchers (at various academic levels) of the cell cycle regulatory process. Three major components working at the Centre, from different backgrounds (computer are considered: the bioentity itself, its spatial localization, and science, physics, theoretical biology, electronic engineering) its temporal localization. CCO provides a test bed for the and from different countries all over the world. CoSBi is a young development of new approaches and tools necessary to create institute with an average researcher age of 32.33 years. a full fledged knowledgebase that enables deployment of www.cosbi.eu advanced reasoning approaches for knowledge discovery and hypotheses generation. Currently, CCO holds almost 50000 A-11 concepts and more that 20 types of relationships. Ontology files exist for four selected model organisms (H. sapiens, S. cerevisiae, The Helmholtz Alliance on systems biology: Towards a S. pombe, and A. thaliana) containing information on cell cycle comprehensive picture of causes of complex disorders and related proteins and the corresponding genes, protein function, diseases localization and posttranslational modifications and protein- Eufinger, Jan1; Klingmüller, Ursula2; Mewes, Hans-Werner3; protein interactions, but also a single ontology with orthology Wanker, Erich4; Lehmann, Irina5; Zilles, Karl6; Eils, Roland7 relationships. 1Helmholtz Alliance on Systems Biology, DKFZ - Theoretical Conclusions: The CCO knowledgebase has resulted in the Bioinformatics - B080, Heidelberg, Germany; 2German Cancer new paradigm of Semantic Systems Biology, a merging of Research Center (DKFZ), A150 - Systems Biology of Signal semantic web technologies and systems biology. We have laid Transduction, Heidelberg, Germany; 3Helmholtz Zentrum the foundations to consolidate this knowledge base and extend it München (HMGU), Institute of Bioinformatics and Systems with key European partners in future EU project settings. Biology, Munich, Germany; 4Max Delbrück Center for Molecular Medicine (MDC), Mechanisms of Neurodegenerative Diseases, A-09 Berlin-Buch, Germany; 5Helmholtz Centre for Environmental Research (UFZ), Department of Environmental Immunology, Text mining for systems biology at the national centre for Leipzig, Germany; 6Research Center Jülich, Institute of Medicine text mining (IME), Jülich, Germany; 7German Cancer Research Center (DKFZ) Ananiadou, Sophia and University of Heidelberg (BioQuant), B080 - Theoretical National Centre for Text Mining, Manchester, United Kingdom Bioinformatics and IPMB, Heidelberg, Germany

With the overwhelming amount of literature, one of the defining The Helmholtz Alliance on Systems Biology was initiated challenges faced by scientists is dealing with information overload in 2007 as a centrally funded initiative comprising a network and information overlook. Without sophisticated new tools, of several Helmholtz-centers, Universities and further external researchers are unable to keep abreast of developments in their partners. The Helmholtz Association, the largest research field and valuable new sources of research data will be under- organization in Germany, has allocated funds of about 24 exploited. The capability of text mining (TM) to find knowledge million Euros until 2011, with the participating Helmholtz centers hidden in text and to present it in a concise form makes it an investing an almost equal sum, to further develop the existing essential part of any strategy for addressing these problems. The capacities in the field of systems biology. UK National Centre for Text Mining (NaCTeM) is the first publicly Scientific focus of the alliance is the elucidation of complex funded text mining centre in the world focusing on providing bio- disease mechanisms using a highly connected and text mining services and tools to the academic community. It also interdisciplinary approach to provide a better understanding plays a critical role in ensuring that researchers are aware of and of how complex diseases such as cancer or diseases of the have access to effective TM solutions. cardiovascular and nervous systems develop. The aim of Text mining techniques enrich the scientific literature and add the alliance is to use systemic approaches to promote the meaning to text by automatically extracting semantic entities understanding of the causes of complex diseases and to develop

46 ICSB 2008 new strategies for treating them. A-13 Research Topics Include: - Signal transmission processes in cancer cells - The molecular Kleio: A knowledge-enriched information retrieval system bases of neurodegenerative and cardiovascular diseases - The for biology influence of toxins on cell metabolism - The role of non-coding Nobata, Chikashi1; Rea, Brian1; Okazaki, Naoaki2; Sasaki, Yutaka1; RNA in regulatory networks - Neuronal structure and the modeling Tsuruoka, Yoshimasa1; Tsujii, Jun’ichi2; Ananiadou, Sophia1 of function in the brain 1University of Manchester, School of Computer Science, The alliance will provide training opportunities for young scientists Manchester, United Kingdom; 2University of Tokyo, Department of and technology platforms for all participating partners. Prof. Computer Science, Tokyo, Japan Roland Eils from the German Cancer Research Center (DKFZ) in Heidelberg heads the initiative, which includes the Helmholtz Objective: The rich variety of term variants is a stumbling block Centers in Heidelberg (DKFZ), München (HMGU), Berlin (MDC), for information retrieval (IR) as these many forms have to be Leipzig (UFZ), Jülich (FZJ) and Braunschweig (HZI) in addition to recognised and mapped from text to existing databases. Our numerous universities and other partners. system Kleio addresses this problem by using our text mining Further information is available under http://www.helmholtz.de/ technology for reducing the diversity of term variation. systemsbiology Results: Kleio is an advanced IR system developed at the UK During the ICSB in Göteborg we will give an overview about National Centre for Text Mining (NaCTeM). The system offers the scientific topics covered in the alliance, present recent and textual and metadata searches across MEDLINE and provides upcoming educational activities, announce open positions enhanced searching functionality by leveraging terminology and explain options for PhD programs. We will also arrange management technologies. Kleio draws upon a number of core possibilities to meet scientists from the alliance during the ICSB technologies from the NaCTeM text mining tool kit to enhance and afterwards. automated detection of biologically important terms appearing in text. One of the key components is AcroMine, which recognises A-12 acronyms (e.g. DEAE) and their definitions (e.g. diethylaminoethyl)

from the whole MEDLINE. It also disambiguates isolated Arenas AGRON-OMICS is a plant science consortium enabling acronyms using their context and maps them into corresponding system level studies of leaf growth definitions. This functionality plays a key role in searching large Fiorani, Fabio document collections by allowing users to expand their queries VIB-UGent, Plant Systems Biology, Zwijnaarde, Belgium without losing the specificity of the original query. Kleio is also dealing with the variety of terms for denoting the same concept. Post-genomic research using model organisms is characterized To map these forms (e.g. IL2, IL-2 and Interleukin-2) to biological by renewed efforts to understand biological complexity through databases we use machine learning based term normalisation integrative approaches and system biology. To enable the techniques which reduce term variation (e.g. il2). An advantage comprehensive study of the molecular mechanisms of plant of applying term normalisation is to permit efficient look-up growth and development, EU is dedicating € 12M to support for 5 and to discover ambiguous and variant terms in the resources. years AGRON-OMICS (Arabidopsis GROwth Network integrating The novelty of our work lies in using existing resources to -OMICS technologies www.agron-omics.eu). Our consortium is a automatically learn term variation patterns. partnership of 14 plant European laboratories leaders in their field Conclusions: Kleio now covers the whole MEDLINE abstracts focusing on the molecular mechanisms regulating leaf growth. with the above technologies. With these technologies KLEIO Objective: The project is built on 7 integrated scientific work- enables query expansion with synonyms in an efficient way. packages and aims to: (i) survey systematically the molecular components driving growth; (ii) explain and classify leaf growth A-14 phenotypes at the molecular level; (iii) build coordinated molecular networks between pathway modules. The netbuilder project - modelling, simulating, and creating Results: The experimental plan combines molecular profiling genetic regulatory networks techniques including the characterization of transcriptome, Wegner, Katja; Schilstra, Maria J. proteome and metabolome in conjunction with automated high- University of Hertfordshire, Biocomputation Research Group, throughput phenotyping. These analyses are conducted on the STRI, Hatfield, United Kingdom same leaf tissue samples grown in controlled environmental conditions. In parallel, our consortium aims to pave the way to the Genetic regulatory networks (GRNs) control intracellular development of novel assays and high-throughput methods for processes by sampling the internal dynamics of a cell, and the functional analysis of plant genes, including high-content cell integrating those along with any external signals from its assays. The large amount of data generated requires dedicated environment and surrounding cell. GRNs control, for instance, data repositories and an appropriate information layer to ensure the development from embryo to adult multicellular organism, access, analysis, and visualization of data sets. To achieve development of intracellular organelles, metabolism, whereas this, the driving principles are to contribute to the improvement many heritable disorders are caused by their malfunctioning. of existing standards and ontologies for data and metadata GRNs are typically very complex and detailed knowledge is still description. to develop robust methods for data integration. scarce. Modelling and simulation techniques play an important Linked to the data integration framework is the construction of role in the understanding and predicting the dynamics of the analytical, mathematical and visualization tools for functional biological processes that underlie the functioning of GRNs, and genomics data intended for building models of complex systems, will also be necessary to exploit their regulatory principles. driven by biological questions. In turn, this computational scaffold We propose to demonstrate our open-source software tool is intended to shape emergent hypothesis to be tested and NetBuilder’, which is being developed to ease modeling and validated experimentally. simulation of GRNs. Modeling and simulation require certain Conclusion:The project will yield data, tools, public resources formal skills that many researchers, particularly those in the field and novel technologies for the plant research community. of experimental molecular and cell biology, cannot be expected to have mastered. To help these researchers understand the structure of biochemical reaction network models, NetBuilder’ implements the Petri net formalism, an intuitive way to model such processes. NetBuilder’ provides an easy to use graphical interface for the creation and visualization of relatively simple networks, as well as a more involved, but more powerful scripting interface for creating larger and more intricate ones. The Petri-net models are automatically translated into standard mathematical models (rate equations, stoichiometry matrix, initial conditions,

ICSB 2008 47 etc.), which can then be simulated using default or user defined development of relevant experimental as well computational parameters and conditions. NetBuilder’ offers a deterministic tools will be an important goal of YSBN. Furthermore, YSBN will and a stochastic simulator, and can also handle mixed models take an active role in training and education, as well as ensure (models that have stochastic as well as deterministic aspects). dissemination of systems biology activities to both the scientific NetBuilder’s modular design includes a model designer, a community and the society. simulation engine and a data handler managing data exchange and plotting functionality. Its design encapsulates and hides many A-17 of the complexities of the modelling and simulation processes from the user. UNICELLSYS - Eukaryotic unicellular organism biology Project web page: http://strc.herts.ac.uk/bio/maria/Apostrophe/ - systems biology of the control of cell growth and proliferation A-15 Hohmann, Stefan1; Markström, Martin2; The UNICELLSYS Consortium, EC-funded2 Systems biology at the institute for advanced biosciences, 1Gothenburg University, Cell and Molecular Biology, Goeteborg, keio university Sweden; 2University of Gothenburg, Goeteborg, Sweden Robert, Martin Keio University, Institute for Advanced Biosciences, Tsuruoka, The overall objective of UNICELLSYS is a quantitative Japan understanding of fundamental characteristics of eukaryotic unicellular organism biology: how cell growth and proliferation are The Institute for Advanced Biosciences (IAB), Keio University, controlled and coordinated by extracellular and intrinsic stimuli. is an academic research institute pioneering the emerging field Achieving an understanding of the principles according to which

Arenas of systems biology, using both experimental and computational bio-molecular systems function requires integrating quantitative biology approaches. Experimental facilities are located in experimentation with simulations of dynamic mathematical the Tsuruoka Town Campus (TTCK) in Yamagata prefecture, models. UNICELLSYS brings together a consortium of leading northern Japan, while the Shonan-Fujisawa campus (SFC), European experimental and computational systems biologists in the greater Tokyo area, hosts the bioinformatics laboratory that will study cell growth and proliferation at the levels of cell and most undergraduate curricular activities. Several research population, single cell, cellular network, large-scale dynamic groups work in collaboration at IAB, focusing mainly on microbial systems and functional module. Building computational post-genomics research, genome design and synthetic biology, reconstructions and dynamic models will involve different precise metabolic engineering, proteomics, metabolomics, RNA biology, quantitative measurements as well as complementary approaches bioinformatics, and computational biology. Using cutting- of mathematical modelling. edge technologies, intracellular components can be analyzed A major challenge will be the generation of comprehensive comprehensively to construct computer simulation models of dynamic models of the entire control system of cell growth and living organisms that can find numerous applications in fields such proliferation, which will require integration of smaller sub-models as biomedical, environmental, and agricultural/food science. and reduction of complexity. Implementation of the models will http://www.iab.keio.ac.jp/ allow observing responses to altered growth conditions zooming Representative publications: in seamlessly from populations consisting of cells of different cell 1. Ishii, N. et al. Multiple High-Throughput Analyses Monitor the cycle stage via genome-wide molecular networks, large dynamic Response of E. coli to Perturbations. Science 316, 593-597 (2007). systems to detailed functional modules. Employing computational 2. Soma, A. et al. Permuted tRNA Genes Expressed via a Circular simulations combined with experimentation will allow discovering RNA Intermediate in Cyanidioschyzon merolae. Science 318, 450- new and emerging principles of bio-molecular organisation and 453 (2007). analysing the control mechanisms of cell growth and proliferation. The project will deliver new knowledge on fundamental eukaryotic A-16 biology as well as tools for quantitative experimentation and modelling. Detailed plans for dissemination and exploitation YSBN - The yeast systems biology network will ensure that UNICELLSYS will have major impact on Nielsen, Jens1; The YSBN Consortium, EC-funded2 the development of Systems Biology in Europe ensuring a 1Chalmers University of Technology, Department of Chemical and competitive advantage of Europe in dynamic quantitative Biological Engineering, Goeteborg, Sweden; 2Chalmers University modelling of bio-molecular processes. of Technology, Goeteborg, Sweden A-18 Systems Biology is an emerging field expected to have a major impact on the future of biological and medical research. It aims at FutureSysBio - Tackling the future challenges in systems systems-level understanding of biological processes by employing biology mathematical analyses and computational tools to integrate Hohmann, Stefan1; Markström, Martin2; Nielsen, Jens3; Hiort, the information content obtained in experimental biology. The Catharina3 Yeast Systems Biology Network (YSBN) will use the yeast 1Gothenburg University, Cell and Molecular Biology, Goeteborg, Saccharomyces cerevisiae as a model system to develop the field Sweden; 2University of Gothenburg, Goeteborg, Sweden; and to advance our understanding of the rules and principles of 3Chalmers University of Technology, Goeteborg, Sweden the dynamic operation of cellular systems. YSBN will serve to integrate activities in yeast systems biology. The emerging field of Systems Biology is anticipated to have The Network will be based on a web resource that also provides a major impact on the biosciences, moving biology from a a database of curated data, mathematical models and simulation phenomenological to a predictive science. Such predictive ability tools. This web resource will allow the research community to should allow to accurately foresee the outcome of therapeutic contribute data as well as mathematical models to simulate interventions with individual patients or to optimise industrial biological processes. The community will then, through further bioprocesses more precisely than has been possible before. experimentation, contribute to the recursive refinement of these Therefore, the results of Systems Biology are expected to have models to advance our understanding of the biological process. major impact on treatment and diagnosing diseases, health care The vision is to generate reference models of budding yeast, and the bio-industries. eventually at the level of the whole cell. Developing the research field and ensuring exploitation of its YSBN will facilitate the cooperation between experimental results therefore is of major social and economic interest for and theoretical yeast researchers, and will stimulate the the European Union. The overall objective of FutureSysBio is generation of quantitative data obtained in time courses and to spur and structure the discussion on the development of containing information on the spatial distribution of the individual Systems Biology and thereby provide guidance to stakeholders components of the yeast cell. Defining the requirements for and scientists. Specifically, the objectives of this project are:

48 ICSB 2008 (1) To gather and update the research community on latest A-20 developments in Systems Biology. (2) To inform and guide funding organisations. (3) To inform and guide pharmaceutical The Gothenburg centre for systems biology and bio-industries in Europe on developments and opportunities Hohmann, Stefan1; Nielsen, Jens2; GSSB, all groups3 in Systems Biology and thereby enable well-informed corporate 1University of Gothenburg, Cell and Molecular Biology, Goeteborg, decisions. (4) To inform and guide higher education and education Sweden; 2Chalmers University of Technology, Goeteborg, funders of challenges and opportunities in interdisciplinary Sweden; 3Chalmers, GU, FCC, Goeteborg, Sweden education and training. (5) To inform the media, policy makers and the general public of opportunities, challenges and facts in We are establishing a Systems Biology Centre in Göteborg Systems Biology. consisting of research teams from Göteborg University (Science To achieve these goals the consortium will organise the following Faculty, Sahlgrenska Academy), Chalmers University of highly visible events: (1) The International Conference on Technology and the Fraunhofer Chalmers Centre. The research Systems Biology ICSB2008 in Gothenburg. (2) The International groups affiliated with the Centre will attack research questions of Conference on Systems Biology ICSB2010 in Edinburgh. (3) wider medical, biological and biotechnological relevance, develop Six high-level expert workshops. (4) Two high-level topical technology for research and serve as resource and collaboration conferences. Different types of dissemination activities tailored partner for Swedish and international institutions. for the target audiences will ensure the significant impact that The Systems Biology research and education landscape in FutureBioSys plans to make on the future development of Göteborg has unique potential. The Science Faculty presently Systems Biology. supports three research platforms relevant for Systems Biology (“Quantitative Biology”, “Chemical Biology” and “Theoretical A-19 Biology”). Those platforms deal with experimentation and modelling in molecular cell biology as well as ecology/ SYSTEMSBIOLOGY - interdisciplinary training in systems epidemiology. Sahlgrenska Academy is home of groups in biology preclinical and clinical environments that combine experimentation 1 2

Hohmann, Stefan ; Nerman, Olle ; The SYSTEMSBIOLOGY and modelling in cell biology and physiology of metabolic and Arenas Consortium, EC-funded3 cardiovascular diseases and understanding the development of 1University of Gothenburg, Cell and Molecular Biology, Goeteborg, allergies. Sweden; 2Chalmers University of Technology, Goeteborg, Chalmers has recently recruited a leading scientist, who will Sweden; 3University of Gothenburg, Goeteborg, Sweden establish significant activities in metabolomics, network biology, bioinformatics and bioengineering. In addition, Chalmers The overall goal of this project is to provide advanced training to houses groups in physics, chemistry, mathematics, computer early-stage researchers in Systems Biology. and systems sciences with interest in biological questions. The Systems Biology uses a combination of experimental and Fraunhofer Chalmers Centre for Industrial Mathematics is home to mathematical approaches to achieve a deeper understanding of a group in Systems Biology. the function and operation of pathways, cells, organs and entire Together, it is estimated that around 20 principle investigators organisms. Systems Biology has enormous potential in future already contribute to activities in Göteborg that are relevant medicine and predictable treatment regimes. The field requires a for the Systems Biology Centre. SWEGENE, supported by the generation of scientists trained in experimental and mathematical/ Wallenberg Foundation, has established technology platforms computational biology. in proteomics, transcriptomics, bioimaging and bioinformatics This project unlocks resources and expertise from: (1) that are relevant for Systems Biology and that will be further Goeteborg: Research groups at Goeteborg University and developed to excellently contribute to generating the data Chalmers University of Technology being part of the”National needed for Systems Biology. Both Chalmers and the Science Research School for Genomics and Bioinformatics”. Faculty presently run master’s programmes in Systems Biology, (2) Berlin: The Graduate School “Dynamics and Evolution of Bioinformatics and Biotechnology. Those will be further developed Cellular and Macromolecular Processes” and the “Berlin Centre of and coordinated. Genome Based Bioinformatics”, with groups from the Humboldt University Berlin, the Max-Planck Institute for Molecular Genetics A-21 plus additional centres. (3) Amsterdam: The Research School “BioCentrum Amsterdam”, The centre for systems biology at edinburgh a joint venture of the Universiteit van Amsterdam (the Kerr, Lorraine; Lebedeva, Galina; Elliot, Elizabeth Swammerdam Institute for Life Sciences, Faculty of Science) and The University of Edinburgh, CSBE, Edinburgh, United Kingdom the Free University of Amsterdam. This EST project relies on the following main activities: (1) The Centre for Systems Biology at Edinburgh, CSBE, is a Centre Five joint PhD projects, in which students undergo their entire for Integrative Systems Biology funded by the BBSRC and PhD programme in the project. EPSRC and led by co-Directors, Professors Andrew Millar and (2) A programme of twenty shorter stays for PhD students. Igor Goryanin. Modelling is central to systems biology, taking (3) Joint courses and workshops, which will be a central part existing knowledge and through static and kinetic models, using of the training, communication, networking and dissemination it to generate new knowledge. The key feature of CSBE is that activities. we place this process at the centre of our endeavour. Through (4) Opening up ongoing local courses to students enrolled at the the work of the Centre, we aim to make theoretical and practical different locations. developments to support all stages of the modelling process. We The EST project will contribute to structuring the field of Systems are establishing an informatics infrastructure that links diverse Biology in Europe such that its scientific and socio-economic data types seamlessly to the tools required for modelling; the potential can be used to address important goals of European Systems Biology Software Infrastructure, SBSI. The modular policies in the health and life sciences. With support from the EC structure of SBSI encompasses varied model construction the project can mobilise resources from national and regional methodology, as well as a suite of tools for model analysis funding bodies to make Europe an attractive place for pursuing a including both static and kinetic analyses. Such large-scale research carrier. modelling frameworks have not previously been available in the academic setting. This work is being informed by three biological exemplar projects which have been chosen to exercise all aspects of the modelling process: the circadian clock in plants, RNA metabolism in yeast and interferon signalling in human macrophages. These projects represent a spectrum of timescales, dynamics, numbers of biological components and overall system complexity.

ICSB 2008 49 The Kinetic Parameter Facility quantifies the rate and affinity descriptions of pathway activation/deactivation in yeast and constants required to parameterise kinetic models, focussing mammalian cells. The approach will be highly systematic. A on the assembly of protein-protein and protein-nucleic acid first network model will be generated based on existing data. A complexes, protein and RNA degradation and the acquisition of reference dataset, including diverse data such as relative and high-quality RNA timeseries. Information about the people and the absolute protein, mRNA and metabolite levels and capturing projects can be found on our regularly updated website http:// the dynamics of the pathway, will be used to establish kinetic csbe.bio.ed.ac.uk. CSBE has members spanning multiple models of yeast and mammalian AMPK. These models will be research institutions and representing diverse areas of research improved and optimised iteratively in several rounds. A major expertise. We also have strong links to industry via direct research challenge will be the attempt to use information from the yeast partnerships, consultancy and our International Science Advisory model to fill gaps in the model describing mammalian AMPK. Board and would welcome further opportunities for collaboration The computational models will support drug development at an from both the public and private sectors. Our philosophy is SME that is fully integrated into all project activities. If successful, strongly collaborative; open source and open access. the results will have major payoff to tackle some of the most rapidly advancing diseases in the modern world, obesity and A-22 type-2 diabetes. Hence, the project will result in a case study for employing systems biology in drug target identification and Systems biology strategies and metabolome engineering in drug development and it will produce results exploitable also for the enhanced production of recombinant proteins in for engineering of microbial metabolism and systems biology Streptomyces - FP6 project Streptomics (www.streptomics. software development. org) Anné, Jozef A-24

Arenas Katholieke Universiteit Leuven, Microbiology and Immunology, division Bacteriology, Leuven, Belgium CELLCOMPUT - biological computation built on cell communication systems Objectives: STREPTOMICS uses Streptomyces as a protein Hohmann, Stefan1; Nordlander, Bodil2; The CELLCOMPUT production platform to study systems biology strategies and Consortium, EC-Funded2 metabolome engineering for the enhanced production of 1University of Gothenburg, Cell and Molecular Biology, Goeteborg, recombinant proteins. It combines genomic data, metabolic Sweden; 2University of Gothenburg, Goeteborg, Sweden network modelling, metabolic flux analysis and proteomics to enhance production of recombinant proteins in Streptomyces. Cell communication systems that form building blocks for This systems biology approach will identify genes/proteins with biological computation devices have significant potential. They key roles in protein secretion and their interrelationship with might be able within the human body to detect and respond to cell growth, secretion stress control and energy production/ changes in the state of health and help to combat disease at consumption. Based on this information a ‘toolbox’ of strains an early stage or serve as sensors for detection of pollutants will be appropriately engineered via deletion of a set of selected coupled to the ability to degrade dangerous substances. genes, and muteins, either via direct mutation of specific amino CELLCOMPUT aims at providing proof of principle that complex acids, or by directed evolution, such that they optimally over- devices consisting of two, three or more programmed cells secrete during fermentation. To obtain this goal, efforts of 4 SMEs can be designed and constructed and at generating building (BioXpr, Direvo, Eurogentec, Prokaria), one major pharmaceutical blocks for such devices. The project will design in silico cells that company (GSK) and 5 academia groups have been combined. have the ability to communicate in predictable manner to form Results: the obtained deliverables will permit the targeted communication systems. The required cells will be constructed manipulation of specific metabolic pathways and the modulation making use of re-engineered yeast signalling pathways - the of key proteins of the protein secretion pathway. As a result of pheromone response and osmosensing HOG pathway - for better understanding of the metabolome-secretome interplay which certain synthetic biology building blocks have already strategies for improved protein secretion will be designed. been prepared. Re-engineering of the pathways will make use These will allow to combine better energy generation and of systems biology approaches established in ongoing projects. directed energy consumption for either cell mass production or The systems will be constructed in yeast cells, improved in heterologous protein secretion. recursive manner integrating simulation of computational models Conclusions: Successful realization of the objectives of the and experimentation and their function will be demonstrated. project will provide useful and commercially valuable innovation to In addition, the project will deliver approaches for systems the EU Streptomyces protein production platform. reprogramming using highly specific mutations or chemical inhibitors. CELLCOMPUT aims at taking part in the development A-23 of synthetic biology in Europe through its research and training activities as well as its dissemination programme. AMPKIN - systems biology of the AMP-activated protein kinase pathway A-25 Hohmann, Stefan1; Elbing, Karin2; The AMPKIN consortium, EC- funded2 Computational systems biology of cell signalling 1University of Gothenburg, Cell and Molecular Biology, Goeteborg, (COSBICS) Sweden; 2University of Gothenburg, Goeteborg, Sweden Wolkenhauer, Olaf; Vera Gonzales, Julio University of Rostock, Systems Biology and Bioinformatics, Biology is moving from describing phenomena to understanding Rostock, Germany design principles and dynamic operation of cellular modules, entire cells, and even organisms. The quantitative approach Objective:The area of cell signalling investigates the transmission to biological systems is driven by technological advances and of information from receptors to gene activation by means of close collaboration between different disciplines. Understanding biochemical reaction pathways. COSBICS is to establish and properties of biological systems holds significant promises apply a novel computational framework in which to investigate for drug development, treatment of diseases or improving cellular signalling. Instead of simply mapping proteins in a bioprocesses. In this project, experimental and theoretical pathway, COSBICS is concerned with “dynamic pathway studies will be integrated to achieve a better understanding of the modelling”. Dynamic pathway modelling establishes mathematical dynamic operation of the AMP-activated protein kinase (AMPK) models to quantitatively predict the spatial-temporal response of signalling pathway. signalling pathways. This pathway plays a central role in monitoring the cellular energy The aims of the project were: I) to obtain predictive dynamic status and controlling energy production and consumption. The models of JAK2-STAT5, Ras/Raf1/MKE/ERK and NFκB pathway, conceptional project idea is to generate kinetic mathematical three signalling pathways commonly subverted in cancer; II)

50 ICSB 2008 the implementation of methodologies to support design of Stellenbosch University (http://jjj.biochem.sun.ac.za); the Vrije experiments; III) the investigation of cross-talk and its coordination Universiteit in Amsterdam (http://jjj.bio.vu.nl); and at Manchester via mathematical modelling and analysis; IV) the investigation of University (http://jjj.mib.ac.uk), with a fourth server being set the role of space and time in the dynamics of signalling pathways. up at CalTech. JWS Online is: 1) a curated model database; 2) Results and Conclusions:Kinetic models have been set up an easy to use simulator (run the models in your browser or via to describe the JAK2-STAT5, RAS/RAF1/MEK/ERK and NFκB webservices); and 3) a Systems Biology service for scientific pathway under the biological conditions investigated in the journals and research projects. JWS Online is linked to four project. A plethora of quantitative experimental techniques have scientific journals for the reviewing of mathematical models: been tested and adapted for the purpose of modelling these FEBS J, IET Systems Biology, Microbiology, and Metabolomics, pathways, including quantitative western blots, life cell imaging and JWS Online is used as a tool in several research projects: and ELISA kits. The Silicon Cell initiative, BioSym, HepatoSys, SysMO, and From the methodological perspective, theoretical and UniCellSys. In addition JWS Online works in close collaboration experimental techniques to investigate protein gradients with Biomodels at EBI. and diffusion effects in the JAK2/STAT5 pathway have been PySCeS is a stand alone application that has been designed designed, implemented and tested. A strategy for the reduction to be functional, accessible and flexible. Using Python allows of dimensionality in non-linear kinetic models of cellular signalling PySCeS to run on a wide variety of operating systems. This, was designed and applied to the investigated pathways. Further coupled with a modular, NoGUI design allow it to be used more a strategy to use kinetic models based on power-law terms interactively or as a module in user developed scripts. PySCeS in cell signalling modelling has been implemented and applied and allows for the analysis of a models structural, simulation, steady a MATLAB toolbox for parameter estimation and optimal design state, stability, control and regulation properties and supports of experiments has been developed and implemented. SBML import/export. New extension modules include: Kraken (grid) and SymCA (symbolic MCA). A-26 Conclusions: The Stellenbosch Systems Biology Centre, headed by the Triple-J group and part of the South Africa National

Professor Kwang-Hyun Cho’s systems biology group in Bioinformatics Network has been developing Systems Biology Arenas Korea since 1999 (Website: http://sbie.kaist.ac.kr) software tools since 2000. Two important packages have been Kim, Tae-Hwan developed thus far: JWS Online and PySCeS. Korea Advanced Institute of Science and Technology (KAIST), Department of Bio and Brain Engineering, Daejeon, Republic of A-28 Korea SysMo–SUMO: Systems understanding of microbial oxygen Professor Kwang-Hyun Cho’s systems biology group in Korea responses was founded in 1999 and currently housed in the Department SysMO, SUMO Consortium1; Ederer, Michael2 of Bio and Brain Engineering at the Korea Advanced Institute of 1www.sysmo.net, Sheffield, Amsterdam, Stuttgart, Magdeburg, Science and Technology (KAIST) as a ‘Laboratory for Systems United Kingdom; 2Max Planck Institute for Dynamics of Complex Biology and Bio-Inspired Engineering (SBIE)’. Professor Technical Systems, Magdeburg, Germany Kwang-Hyun Cho’s group has been working on systems-level investigations of cellular signal transduction pathways, reverse SysMO: SUMO (Systems Understanding of Microbial Oxygen engineering of biomolecular regulatory networks, and unraveling Responses) is a project within the transnational SysMO initiative hidden cellular dynamics. The focus has been on developing a (Systems Biology of Microorganisms, http://www.sysmo.net, systems biology analysis of cellular information processing by project 3). signaling and gene networks in cells with particular emphasis Description of project: The SUMO consortium investigates towards understanding cell-fate decisions on proliferation and the responses of Escherichia coli to oxygen availability by differentiation. We have two long term objectives in our research. systems biological methods. We collectively obtain and analyze, The first is to create a predictive model for a programmable cell using commonly agreed protocols, transcriptomic, proteomic, that can be optimized for personalized therapy, which has been metabolomic and biochemical data sets that describe the already underway for over 10 years. Our second objective is dynamics of the response to oxygen. Data sets are integrated to longer term and is to apply the knowledge obtained from the elaborate predictive mathematical and computer science models study of molecular biological systems to engineering, which we in an iterative process of model-based hypothesis generation and call as bio-inspired engineering. In this way we hope to contribute experimental design. to engineering innovation using ideas inspired by molecular Specifically, we investigate how this bacterium senses oxygen, systems biology. At the arena of ICSB2008, current research or the associated changes in oxidation/reduction balance, via focuses and some of on-going funded research projects of the Fnr and ArcA proteins, how these systems interact with Professor Kwang-Hyun Cho’s group are to be presented. other regulatory systems, and how the redox response of an E. coli population is generated from the responses of single cells. A-27 There are four sub-projects to determine system properties and behavior and three sub-projects to employ different and Systems biology tools out of Africa: JWS online and complementary modeling approaches using published data sets PySCeS and data emerging from our own work. We construct increasingly Snoep, Jacky1; Olivier, Brett2; Van Gend, Carel2; Du Preez, elaborate models of the system at different levels of detail, which Franco2; Conradie, Riaan2; Akhurst, Tim2; Dominy, James2; are used to generate new hypotheses and influence further Penkler, Gerald2; Holm, Kora2; Rohwer, Johann2; Hofmeyr, experimental design. Jannie2; Stoof, Cor3; Westerhoff, Hans3 Project partners: The SUMO consortium consists of projects 1University of Amsterdam, Department of Biochemistry, located at the University of Sheffield (Prof Robert Poole and Stellenbosch, South Africa; 2Stellenbosch University, Department Prof Jeff Green from the Department of Molecular Biology of Biochemistry, Stellenbosch, South Africa; 3Vrije Universiteit, and Biotechnology, Prof Mike Holcombe from the Deparment Molecular Cell Physiology, Amsterdam, Netherlands of Computer Science), at the Swammerdam Institute for Life Sciences of the Universiteit van Amsterdam (Prof Joost Teixeira Motivation: With the explosive growth of the Systems Biology de Mattos), at the Institute for System Dynamics of the Universität discipline in the last decade it has become increasingly clear Stuttgart (Dr Thomas Sauter, Prof Oliver Sawodny) and at the that good software tools are essential to move the field forward. Max Planck Institute for Dynamics of Complex Technical Systems Good tools for the analysis, comparison, and storage of kinetic in Magdeburg (Dr Katja Bettenbrock, Prof Ernst Dieter Gilles). models and for the linking of experimental data to these models Project leader is Prof Robert Poole ([email protected]). are needed. Results: JWS Online is a server based tool, accessible at:

ICSB 2008 51 A-29 Objective: BaSysBio adopts a systems biology approach in which quantitative experimental data will be generated for each The centre for plant integrative biology, university of step of the information flow and will fuel computational modelling. Nottingham High throughput technologies (living cell arrays, tiling DNA Holman, Tara; Hodgman, Charlie; Middleton, Alistair; Lydon, microarrays, MDLC proteomics and quantitative metabolomics) Susannah will be developed in conjunction with new computational University of Nottingham, Centre for Plant Integrative Biology, modelling concepts to facilitate the understanding of biological Nottingham, United Kingdom complexity. Models will simulate the cellular transcriptional responses to environmental changes and their impact on The Centre for Plant Integrative Biology (CPIB), based at the metabolism and proteome dynamics. The iterative process of Sutton Bonington Campus of the University of Nottingham, is one simulations and model-driven targeted experiments will generate of six flagship Centres for Integrative Systems Biology in the UK, novel hypotheses about the mechanistic nature of dynamic funded by the BBSRC and EPSRC. cellular responses, unravel emerging systems properties and CPIB aims to create a multiscale model of a plant root, based on ultimately provide an efficient roadmap to tackle novel, pathogenic the model higher plant Arabidopsis thaliana. Modelling the root organisms. is essential to improving our understanding of plant growth and This system-based strategy will enable BaSysBio i) to understand development, and its response to environmental factors. Use of how transcriptional regulation and metabolism are quantitatively Arabidopsis will aid eventual transfer to crop species, to address integrated at a global level; ii) to unravel cellular transcriptional real-world challenges in food production. responses in conditions mimicking pathogenesis. Finally, the The model will be capable of integration with the virtual shoot project will validate the general applicability of the knowledge being developed in the ComputablePlant project, paving the and integrated modelling-experimental strategy developed in the

Arenas way for an integrated model plant. The development of a ‘virtual highly tractable Bacillus subtilis model towards an understanding root’ will also serve as an exemplar for using Integrative Systems of regulatory networks controlling pathogenesis in disease- Biology in a multicellular system. causing bacteria. The five year project comprises four research strands: Results: After 18 months, the consortium uses the technology • Strand 1 focuses on cell elongation during radicle emergence and experimental methods to generate data with B. subtilis. A first and primary root growth large set of transcriptomics, proteomics and metabolomics data • Strand 2 focuses on cell division in the root apical meristem are being generated from samples from the same nutrient-shift • Strand 3 (commencing in 2009) will examine the initiation, experiment. The technology is becoming ready for B. anthracis patterning and emergence of lateral roots and S. aureus. Large-scale stoichiometric network models for B. • Strand 4 will integrate the models at different physical scales subtilis metabolism as well as computational methods for network (molecules to whole organ) across the first three strands reconstruction and analysis have been established. Standard Expertise at CPIB includes: relational databases for large scale experimental data have been • Transcriptomics experimentation and analysis implemented. Standard Operating Procedures are developed in • Mathematical (ODE) modelling of hormone networks a reference lab, and validated in different labs before use by the • Experimentation on hormone-related behaviour at the cell level consortium. • Localisation using antibodies raised against root proteins Conclusions: The computational/experimental SB loop is gaining • Novel imaging techniques for plant growth momentum. (www.basysbio.eu) • Biomechanical exploration of plant structures and properties • Mathematical modelling of biomechanical properties A-31 • Finite Element simulation and modelling • Multiscale asymptotic modelling techniques The MPIMG medical systems biology approach - a new • Model optimisation path in combating human diseases • Bioinformatics underpinning the above Wierling, Christoph; Maschke-Dutz, Elisabeth; Kamburov, Atanas; • Outreach and training development Daskalaki, Andriani; Lehrach, Hans; Herwig, Ralf CPIB welcomes enquiries from potential collaborators, fellowship Max Planck Institute for Molecular Genetics, Berlin, Germany holders and students. CPIB website: http://www.cpib.info Objective: The Max Planck Institute for Molecular Genetics (MPIMG,http://www.molgen.mpg.de) is one of the core centres A-30 of genome research in Europe. Several important projects have been carried out in functional genomics, proteomics and systems BaSysBio: Towards an understanding of dynamic biology (for example the sequencing of human chromosome transcriptional regulation at global scale in bacteria: a 21) that build the basis for applying genomic tools to medical systems biology approach problems. Sauer, Uwe1; Jung, Alexander2; Molina, Franck3; Aymerich, Medical systems biology has recently been identified as a Stephane4; Banga, Julio R.5; Hecker, Michael6; Stelling, Joerg7; primary research goal because of its (yet unproven!) potentials for Bessières, Philippe4; Fischer, Hans-Peter8; Schwikowski, Benno9; accelerating drug development and improving therapies with the Msadek, Tarek9; Fromion, Vincent4; Sautot, Caroline10; Klipp, use of computer models. The development of disease models is Edda11; Van Dijl, Jan Maarten12; Devine, Kevin13; Jarmer, Hanne a trade-off between model size and prediction precision. Models Ø.14; Harwood, Colin15; Wilkinson, Anthony16; Lewis, Peter J.17; with high precision generate computational predictions of large Noirot, Philippe4 detail based on a rather small number of model components, 1ETH Zurich, Institute of Molecular Systems Biology, Zurich, a strategy currently followed in the majority of cases. Models of Switzerland; 2Applera Deutschland GmbH, Ueberlingen, large size have higher generality by covering many disease genes Germany; 3CNRS UMR 5160, Montpellier, France; 4I.N.R.A., Jouy but their sizes have implications on the depth of computational en Josas, France; 5C.S.I.C., Vigo, Spain; 6Ernst-Moritz Arndt- predictions. All predictions derived from kinetic models are, Universität Greifswald, Greifswald, Germany; 7ETH Zurich, Zurich, however, often compromised by the difficulties to measure the Switzerland; 8Genedata, Basel, Switzerland; 9Institut Pasteur, relevant parameters under in-vivo conditions. The different steps Paris, France; 10I.N.R.A. Transfert, Paris, France; 11Max Planck in the generation of disease models are (a) the identification of Institute for Molecular Genetics, Berlin, Germany; 12University predictive markers with respect to the disease domain, (b) the Medical Center Groningen, Groningen, Netherlands; 13Trinity annotation of reaction systems relevant for the disease, (c) the College Dublin, Dublin, Ireland; 14The Technical University of generation of computational predictions derived from in silico Denmark, Lyngby, Denmark; 15Newcastle University Medical perturbations of these networks. School, Newcastle upon Tyne, United Kingdom; 16University of Results and Conclusions: In our arena we present practical York, York, United Kingdom; 17University of Newcastle, Callaghan, results that address these three steps. Results have been carried Australia out within several projects involving medical systems biology, in

52 ICSB 2008 particular the EU projects EMI-CD, ESBIC-D, SysProt, EMBRACE differentiation, endocytosis, detoxification and iron metabolism in (all Framework 6) and APO-SYS (Framework 7) with focus on liver cells. To this end, experimental research teams from biology, cancer and type-2 diabetes mellitus. We describe the strategies chemistry, pharmacology and medicine work hand in hand with for developing disease models in these different projects. representatives of theoretical physics and mathematics as well as Furthermore, several of our software tools, such as the modelling with computer scientists and engineers. and simulation system PyBioS and the pathway integration With the foundation of FORSYS - Research Units of Systems system CPDB are presented by computer demonstrations. Biology in 2007, the BMBF aimed to expand and strengthen on-topic Systems Biology research in Germany. Its main objective A-32 is to improve the existing research infrastructure and thus create an ideal basis for the further development of this pioneering RIKEN advanced science institute, advanced research branch. Four centres have received funding for five computational sciences department years. They are integrated into the infrastructure of the local Onami, Shuichi universities and research institutions. In addition to establishing RIKEN Advanced Science Institute, Computational Systems junior research groups, attention will also be devoted to educating Biology Group, Yokohama, Japan and training students and Ph.D. candidates. The Freiburg Initiative for Systems Biology (FRISYS) works on “Signalling in Growth and Our research objective is the establishment of RIKEN as a major Differentiation” to investigate specific aspects of cellular signal center of computational science in the 21st century, the age transduction in relation to the growth and differentiation of cells. of prediction. We focus on computational sciences for the life The Heidelberg VIROQUANT project studies the systems biology sciences, which have grown greatly in importance at RIKEN of virus-cell interactions. The focus of Magdeburg’s Centre for in recent years, incorporating materials and mathematical Systems Biology (MaCS) is on the development of new systems sciences, which have key importance to create a novel foundation biology approaches and their application for the analysis and for development of computational sciences. RIKEN’s Next- reconstruction of molecular networks of signal transduction and Generation Super Computer is planned as a platform to maximize regulation on a cellular level. The objective of GoFORSYS in

exploitation, and as a core organ of its research effort. Potsdam is to achieve a better understanding of photosynthesis, Arenas Computational Systems Biology Research Group its regulation in response to selected environmental factors, and The goal of this group is to understand biological systems by the resulting influences on cell growth. efficiently computing molecular interactions and to develop new drug discovery strategies by taking advantage of systems- A-34 level understanding of biological systems. We also promote computational researches that are closely related to experimental The systems biology network at Milano-Bicocca data, such as bioinformatics, computational systems biology, and Alberghina, Lilia computational genomics, and establish computational synthetic Università di Milano-Bicocca, Milano, Italy biology and computational molecular design. - Makoto Taiji: High-Performance Molecular Simulation Two departments (Biotechnology and Biosciences and - Mariko Hatakeyama: Cellular Systems Modeling Informatics, Systems and Communication) of the University of - Todd Taylor: MetaSystems Modeling Milano-Bicocca, Milan, Italy, are developing together systems - Pawan K. Dhar: Synthetic Biology biology projects, the main focus being on control and execution - Shuichi Onami: Developmental Systems Modeling of cell cycle and apoptosis, both in budding yeast and in Living Matter Simulation Research Group mammalian cells. A modular systems biology approach has been We started the computational bio-engineering simulation project set up that relies on molecular analysis (using both post-genomic from 1999. From 1999 to 2004, we had studied in three fields: and conventional techniques) of perturbed cells for network circulatory system simulation, hard/soft tissue simulation and identification and parameter estimation. In collaboration with body motion simulation. We had succeeded in developing modelers, in Italy, Germany and USA, models of yeast cell cycle simulation software and measurement systems and could at various level of definition have been developed for cells and for continue the project. Currently, we are developing human populations. The major achievement of this line of work has been segmented models with 1 mm resolution and also software to the recognition of the critical cell size required to enter S phase simulate blood flows and deformation of soft and hard tissues. as an emergent property of the newly defined G1/S network. - Ryutaro Himeno: Living Matter Simulation Molecular analysis of cell cycle modulation by nutrients is used - Hideo Yokota: Live Cell Modeling to better define the structure of the network and its connections Graduate Student and Postdoctoral Researcher positions are to major signaling pathways. A model of the Ras pathway has available. For more information, please visit our website: http:// already been reported. The next step is to model the initiation of www.riken.jp/asi/en/lab_all.html#acsd. Questions can be directed DNA replication, the following step in the cycle. to Shuichi Onami ([email protected]). Parallel effort is directed to the study of the cell cycle and of its connection with other cellular pathways in normal and A-33 transformed fibroblasts. Definitions of molecular pathways modified by transformation is being undertaken and the The German systems biology activities FORSYS and development of a mathematical model of the G1/S transition HepatoSys in mammalian cells, that takes into account the nuclear/ Heisner, Ute1; Hollmann, Susanne2; Stebel, Sabine3 cytoplasmic localization of cell cycle players, is under way.Time 1Institute for Physics, University Freiburg, Freiburg, Germany; series experiments of transcriptome analysis are better defining 2Coordination Office Magdeburg Center for Systems Biology response of normal and transformed cells to nutrients with a (MaCS), Otto-von-Guericke Universität, Magdeburg, Germany; bioinformatic analysis, lines performed in collaboration with 3FRISYS Project Administration, Albert-Ludwigs-Universität the National Research Council Institutes in Milano and Milano- Freiburg, Freiburg, Germany Politecnico. Finally a line of work develops models of apoptosis in neurons. Systems Biology requires new research structures in science Long term aim of these studies is to support drug discovery, in and industry, new cooperation models and a new quality of collaboration with pharma/biotech industries. interdisciplinary and inter-industrial collaboration in a national and international framework. The Federal Ministry of Education and Research (BMBF) recognized these needs early on and has acted accordingly. The competence network HepatoSys was founded in 2004 as the first Systems Biology network in Germany. With more than 40 working groups all over Germany it investigates regeneration,

ICSB 2008 53 A-35 ELIXIR will include a transnational biological information service, support for core and specialised data resources, a European Oxford centre for integratives systems biology and systems biomolecular data centre, infrastructure for biological information biology doctoral training centre in the new member states, training to optimise use of the data. Jones, Nick1; Wadhams, George2; Smith, David3; Novak, Bela4 ELIXIR will contribute to European science by, optimising access 1Oxford Physics/Biochemistry, Oxford, United Kingdom; 2Oxford to and exploitation of life-science data, ensuring longevity of data Biochemistry, Oxford, United Kingdom; 3Oxford Mathematics, and protecting investments already made in research, increasing Oxford, United Kingdom; 4Oxford Biochemistry, Oxford, United the competence and size of the user community by strengthening Kingdom national efforts in training and outreach, enhancing the global success and influence of Europe in life-science research and Oxford University has established a new centre in Systems industry. Biology. The aim and practice is to couple experimental work in a few A-37 focussed areas with concomitant theoretical support and unifying work by 16+ groups. We believe the best way forward is to pick The computable plant a limited number of tractable and important biological systems Jönsson, Henrik1; Chickarmane, Vijay2; Gordon, Sean2; Heisler, including: the bacterial histidine protein kinase (HPK) dependent Marcus2; Krupinski, Pawel1; Melke, Pontus1; Meyerowitz, Elliot2; chemosensory pathway; a defined section of the eukaryotic cell Mjolsness, Eric3; Roeder, Adrienne2; Sahlin, Patrik1; Shapiro, cycle pathway; the cellular hypoxic response; and the eukaryotic Bruce2 flagellar proteome - and to aim to develop reliable, predictive 1Lund University, Lund, Sweden; 2California Institute of models for those systems. No single model or modelling Technology, Pasadena, United States; 3University of California

Arenas approach will be capable of describing all aspects of a complex Irvine, Irvine, United States biological system, and therefore different modelling approaches will be used for different types of biological data. The computable plant is a multidisciplinary collaboration initiated Theoretical approaches include: Bioinformatics and Data by Eric Mjolsness, UC Irvine, and Elliot Meyerowitz, California Analysis; Molecular Simulations; Continuous, Discrete, Stochastic Institute of Technology, and involves several groups around the and Hybrid Models; Network Theory; Machine Learning and world. Genetic manipulations, live imaging, image processing, Visualisation techniques. and computational modeling are integrated with the aim of an www.sysbio.ox.ac.uk increased understanding of plant shoot development. Several Alongside the new research centre is a new 4-year graduate aspects of the development are addressed where two pursued programme, or Doctoral Training Centre, with first year taught examples are stem cell regulation and organ initiation (see e.g. courses and lab rotations. We aim to produce quantitative PNAS 103, 1633 (2006)). interdisciplinary scientists who can switch naturally between in In this arena we will display several of our models and show silico and in vivo research. The research focus of the Systems related data acquired with confocal microscopy. We will also Biology Doctoral Training Centre will be in developing a systems show some of the software developed within the computable approach to “bridging the gap” between theoretical and plant project with emphasis on tools for modeling multi-cellular experimental knowledge from the level of the individual protein tissue and for extracting cells from three-dimensional confocal to the level of the whole cell / organism, by providing a training data. People will be given the opportunity to simulate and programme across the breadth of relevant physical and life manipulate models for plant shoot development. sciences techniques. There are four broad themes of study More information about the computable plant project can be within the centre: Chemosensory networks and whole organism found at http://www.computableplant.org. behaviour; Signalling Pathway Modelling; Molecules to Cells; Integrative Biology. A-38 www.sysbiodtc.ox.ac.uk Richard D. Berlin center for cell analysis and modeling A-36 Moraru, Ion; Blinov, Michael; Cowan, Ann; Holmes, Raquell; Slepchenko, Boris; Schaff, James; Loew, Leslie Systems biology at the European bioinformatics institute University of Connecticut Health Center, Farmington, United Le Novere, Nicolas States EMBL-EBI, Hinxton, United Kingdom The Richard D. Berlin Center for Cell Analysis and Modeling As we move towards understanding biology at the systems level, (CCAM; http://www.ccam.uchc.edu/) at the University of access to large data sets of many different types has become Connecticut Health Center is a National Technology Center crucial. The European Bioinformatics Institute (EMBL-EBI, http:// for Network and Pathways (http://ntcnp.org/) funded by an www.ebi.ac.uk/), which is part of the European Molecular Biology NIH Roadmap initiative. We focus on the development of Laboratory (EMBL), is one of the few places in the world that new technologies to accelerate discovery and facilitate the has the resources and expertise to fulfil this important task. comprehensive study of biological pathways and networks. EBI’s data resources offer systems biologists access to gene Research and training at CCAM builds on faculty strengths in expression (ArrayExpress), protein identification (PRIDE) and physics, chemistry, experimental biology, live cell imaging, and interactions (IntAct), structure of macromolecules (MSD) and computation. small chemicals (ChEBI), pathways (Reactome) and mathematical CCAM addresses a fundamental problem of cell function: how models (BioModels DB). In addition, the EBI is a major player in signaling networks regulate, and are regulated by, the spatial the development of standards (MIAME, MIRIAM, MIMIx, MIAPE organization of molecules in cells. We integrate new microscope ...), ontologies (GO, ChEBI, IntEnz, SBO, OBI ...) and formal technologies for quantitative in vivo live cell measurements representations (SBML, MAGE-ML, PSI-MI ...), three types of with new physical formulations and computational tools that resources abolutely essential to Integrative Systems Biology. produce spatially realistic simulations of intracellular dynamics. The EBI coordinates the development of the European Life Combining such methods with new techniques to manipulate the Science Infrastructure for biological information (ELIXIR, http:// distribution of molecules in living cells provides an excellent suite www.elixir-europe.org/) which mission is to build a sustainable of computational and experimental tools for understanding cellular support for life science research and its translation to medicine, spatial, temporal dynamics. environment, bio-industries, and society. To this end, Europe Quantitative modeling of cellular systems is the centerpiece of needs coordinated data resources, with improved access and our efforts. Visitors to the CCAM Arena will be introduced to links with data in other related domains; a united European several modeling technologies developed at our center, including voice to influence global decisions and maintain open access; specialized tools for combinatorial complexity and for analysis adequate, sustainable, funding for this distributed infrastructure. and modeling of confocal images. One of the projects hosted at

54 ICSB 2008 CCAM is the Virtual Cell platform (VCell; http://vcell.org/). VCell is Research Institute for Mathematics and Computer Science) and a collaborative, distributed modeling and simulation environment, AMOLF (FOM Institute for Atomic and Molecular Physics). It and is itself an NIH-funded National Resource. A schedule of amalgamates biological and biomedical sciences with chemistry, software demos at the Arena will be available, as well as handouts physics, mathematics, informatics and engineering. of user guides and tutorials. 1. NISB develops and exploits new systems biology tools and CCAM’s graduate training program in quantitative cell biology approaches to investigate complex biological systems and unveil trains students from physics, mathematics and engineering underlying principles. Biological systems are approached as backgrounds alongside students with traditional biology networks of molecules, cell, tissues and organisms that interact in backgrounds. We also offer postdoctoral training, faculty time and space. fellowships and undergraduate summer fellowships. Short 2. NISB also contributes to the development of novel managerial courses in modeling and microscopy occur throughout the year. tools that are required to achieve the scaling up of biomedical Materials on training opportunities will also be available at the and biotechnological research that is essential for effective Arena. application of systems biology approaches in biomedical and biotechnological research. A-39 3. NISB has the lead in the Netherlands Consortium for Systems Biology (NCSB), a national systems biology program (30 M€, 5 SystemsX.ch - the Swiss initiative in systems biology years) of the Netherlands Genomic Initiative (NGI, www.genomics. Vonder Mühll, Daniel1; Aebersold, Ruedi2 nl). 1SystemsX.ch, Zurich, Switzerland; 2ETH Zurich, Institut for 4. NISB has taken the initiative, together with other systems Molecular Systems Biology, Zurich, Switzerland biology institutes in Europe to develop a large-scale pan- European 10 years research program that tackles metabolic Objective: SystemsX.ch is supporting and promoting the syndrome, using systems biology as a driver and integrator: the paradigm shift to a quantitative and predictive biological science Systems Biology to Combat in Switzerland’s academic institutions. Therefore, strengths in 5. Metabolic Syndrome program (SBMS, http://www.esf.org/

genomics and biomedical research are combined with the basic activities/eurobiofund/eurobioforum-2007/sbms.html#c22175). Arenas research in chemistry, mathematics, physics, and engineering The SBMS program is a large-scale pilot to merge biomedical by funding interdisciplinary research projects in a national research and systems biology at a European level. Aim is to start partnership. An important aim is to train and educate the next first projects in 2009. generation of systems biology researchers. In addition, various 6. NISB and its research programs has many open positions in forms of public-private-partnerships by systems biology research the field of systems biology and its applications in biomedical, collaborations between academia and private enterprises are biotechnological and agricultural research. envisaged. SystemsX.ch is a network of large Research, Technology and A-41 Development Projects (RTDs), Interdisciplinary PhD Projects (IPhDs), and Interdisciplinary Pilot Projects (IPPs), which are done Center systems biology (CSB) Stuttgart, Germany at eleven Swiss universities and research institutions. Federal Reuss, Matthias Government has approved CHF 100 Mio for 2008-2011. The Universität Stuttgart, Center Systems Biology, Stuttgart, Germany involved institutions have to match the projects with at least the same amount. Additionally there are supplementary funds coming The newly established Center Systems Biology (CSB) is an inter- from industry or other funding agencies. faculty research institution of the University of Stuttgart with Status Quo August 2008: the mission to establish and sustain new structures that allow After the first Call for Proposal, the Swiss National Science integrated research and education across disciplines in the field of Foundation and SystemsX.ch approved a total of 35 projects: systems biology. 8 RTDs, 15 IPhDs and 12 IPPs. Within the RTDs more than 75 The established, long standing collaboration between modelling reserach groups will work on the following topics: and experimental groups at the University Stuttgart becomes • YeastX - Towards an Understanding of Nutrient Signaling and manifest in the special structure of the individual projects. The Metabolic Operation (PI: Uwe Sauer, ETH Zurich) collaboration between wet and dry lab in each of the projects • DynamiX - Yeast Protein Network Dynamics (PI: Sebastian strengthens important aspects of quantification, strongly linked Maerkl, EPF Lausanne) with topics such as optimal experimental design, iterative cycles • Neurochoice - Neural Correlates of Collective Decision- between model and experiment, model validation, discrimination Making (PI: Fritjof Helmchen, University of Zurich) and identification. • LipidX - Systems Biology of Biomembranes (PI: Gisou van der The present research program, supported by the Ministry of Goot, EPF Lausanne) Science, Research and the Arts Baden-Württemberg and the • PhosphoNetX - Phosphorylation Modulated Networks of the University Stuttgart, is structured into three sections: Cell (PI: Ruedi Aebersold, ETH Zurich) (A) Cellular compartments: Dynamics and signalling processes • LiverX - Systems Biology of Hepatitic Insulin Resistance (PI: (B) Reconstruction and dynamic analysis of molecular networks Wilhelm Krek, ETH Zurich) (C) Systems theoretic methods for modelling and analysis in • WingX - Systems Biology of the Drosophila Wing (PI: Ernst Systems Biology with applications to mammalian signalling and Hafen, ETH Zurich) bacterial metabolism • Plant Growth in a Changing Environment (PI: Cris Kuhlemeier, Two further research programs are coming up University of Bern) A)”A Systems Biology Approach towards Predictive Cancer The 35 SystemsX.ch projects have just been started. A 2nd call for Therapy” proposals will be published in fall 2008. B) “Systems Biology in Pseudomonas for industrial biocatalysis” Furthermore members of the CSB are contributing to A-40 transnational initiatives in Systems Biology, such as five SYSMO consortia, YSBN, Hepatosys, to name a few. Netherlands institute for systems biology (NISB) Van Driel, Roel Netherlands Institute for Systems Biology (NISB), Amsterdam, Netherlands

The Netherlands Institute for Systems Biology (NISB, www.sysbio. nl) combines the efforts in the field of systems biology of 21 research groups of the four NISB partners: the Vrije Universiteit Amsterdam, the University of Amsterdam, the CWI (Netherlands

ICSB 2008 55 A-42 A-44

BioQuant - center for quantitative Analysis of molecular CellDesigner and payao: Constructing and sharing and cellular bio systems - university of heidelberg, biological networks Germany Kitano, Hiroaki Eils, Roland1; Oberthuer, Angela2 The Systems Biology Institute, Okinawa Institute of Science and, 1University of Heidelberg, BioQuant, German Cancer Research Okinawa, Japan Center (DKFZ), Heidelberg, Germany; 2University of Heidelberg, BioQuant, Heidelberg, Germany Constructing and sharing biological network is critical in systems biology research. We have been working of development of BioQuant the recently established Center for “Quantitative bioligcal network construction and sharing tools: CellDesigner and analysis of molecular and cellular bio systems” at the University of Payao. CellDesigner is a tool for constructing biological networks Heidelberg is solely dedicated to research in Systems Biology. and enables simulation and analysis by connecting them to tools It is the core structure of an interdisciplinary research network that provide such services. It is a gateway software, and now that integrates system biology research pursued at various reported to be the most widely used specialized software in research centers at the University of Heidelberg, the Medical systems biology (Klipp, et al., Nature Biotechnology, March 2007). Faculty and University hospitals Heidelberg, the German Cancer Sume of users create a hugh interaction maps. Research Center, the European Molecular Biology Laboratories, One of problems with huge interaction maps is how to up-date the European Media lab and the Max-Planck-Institute for Medical them, and carry out correction and modification processes Research. At the BioQuant Center experimental scientists and in a consistent manner. Payao is a web2.0 style knowledge theoreticians carry out systems biology research projects together assembling and exchanging tool that enable users to post

Arenas under one roof. This ensures a swift validation of scientific biological networks and enable community-based tagging, hypotheses via experimental data and a constant refinement of annotation, and up-date of the network. Users can add note, mathematical models. At present, up to 40 University and non- post debate about interpretation of the map, suggest up-date on University research groups are affiliated to BioQuant. specific properties of maps, and create external links to relevant BioQuant’s central technology platform provides advanced web sites and publications. Payao is now being linked with computational tools and methods for data analysis, visualization PathTEXT text-mining tool by University of Tokyo and University and modeling as well as cutting edge technologies and of Manchester to enable users to search relevant papers using equipment in the area of high-throughput screening and network-based interface. Payao service is hosted at payaologue image acquisition, high-resolution microscopy, single molecule web site. spectroscopy and cryo-EM for the BioQuant research community. The Nikon Imaging Center at the University of Heidelberg is an A-45 integral part of BioQuant’s central technology platform. BioQuant’s research program consists of four distinct but closely iBioSim: A tool for the analysis and design of genetic interconnected project areas dedicated to the investigation circuits of: - protein machines: biogenesis, interaction and regulation; - Myers, Chris J.1; Barker, Nathan2; Kuwahara, Hiroyuki3; Madsen, dynamics of cell architecture; - information processing in complex Curtis4; Nguyen, Nam-Phuong D.4 multi-cellular networks and - alteration of cellular networks 1Dept. of Electrical and Computer Eng., U. of Utah, Salt Lake by infectious pathogens. Currently, research teams of four City, UT, United States; 2Dept. of CSIS, Southern Utah University, prestigious scientific programs are accommodated in BioQuant, Cedar City, UT, United States; 3Microsoft Research - U. of Trento, namely FORSYS-ViroQuant, funded by Federal Ministry for Centre for Computational and Systems Biology, Trento, Italy; Education and Research (BMBF), the public- private-partnership 4School of Computing, U. of Utah, Salt Lake City, United States program BioMS, the excellence cluster CellNetworks funded by the DFG as well as parts of the Helmholtz program Systems iBioSim supports learning of genetic circuit models, efficient Biology of Cancer (SBCancer). abstraction-based analysis of these models, and the design of Further information is available under www.bioquant.uni- synthetic genetic circuits. iBioSim includes project management heidelberg.de features and a graphical user interface that facilitates the development and maintenance of genetic circuit models as well A-43 as both experimental and simulation data. Models in iBioSim can be created using either an SBML editor or a Genetic Circuit Model FACTA: An efficient text search engine for finding (GCM) editor. The SBML editor and the iBioSim simulation engine associations between biomedical concepts support virtually all of SBML Level 2 Version 3 including reactions, Tsuruoka, Yoshimasa1; Tsujii, Jun’ichi2; Ananiadou, Sophia1 rules, events, constraints, etc. The GCM editor improves the 1The University of Manchester, Manchester, United Kingdom; 2The efficiency of model development by supporting modeling at a University of Tokyo, Tokyo, Japan higher level of abstraction than the molecular level supported by SBML. Namely, a GCM includes only important species and their FACTA is a text search engine for mining information about influences upon each other. iBioSim can automatically translate pairwise association between biomedical concepts such as from GCM to SBML models for analysis. A GCM can be either genes/proteins, diseases, enzymes and chemical compounds. manually created or automatically learned from time-series data. The system aims to help researchers by providing a quick and iBioSim also includes an efficient simulation engine that supports interactive way of exploring such information in MEDLINE. Unlike ODE, stochastic, and Markov chain analysis of these models. existing systems that share a similar purpose, FACTA pre- This engine utilizes automatic abstraction to improve analysis indexes not only the words but also the concepts appearing in time, often by one to two orders of magnitude. Finally, iBioSim the documents, which enables the user to immediately receive has a graphical editor for visualizing both time series and event the mining results even when the number of the documents that probability analysis results. match the query is very large. The retrieved concepts are ranked iBioSim has been applied successfully to numerous projects according to their relevance to the query when presented to the including an analysis of the phage λ decision circuit and the E. coli user. The user can also easily view snippets from MEDLINE to get Fim switch. It has also been applied to the design of a synthetic textual evidence of associations. genetic Muller C-element, an asynchronous state-holding gate. In these and other efforts, the iBioSim tool with its support for automatic abstraction has been shown to greatly improve the productivity of researchers who are analyzing and designing genetic circuits. This arena will complement our tutorial on iBioSim by allowing for individual demonstrations and discussions. More information including download access for the iBioSim tool is

56 ICSB 2008 located at: http://www.async.ece.utah.edu/iBioSim/. additional aspects of metabolic regulation and signal transduction including both prokaryotic and eukaryotic systems. A-46 Although the research group itself is composed of researchers from different disciplines, all activities are based on a close The centre for integrated systems biology of ageing and cooperation with external biological groups, with the main focus nutrition (CISBAN) of their work on the specific research area of cooperation. The Shanley, Daryl1; Wilkinson, Darren2; Wipat, Anil3; Lydall, David1; group has built up a well-equipped fermentation laboratory to von Zglinicki, Thomas1; Kirkwood, Tom1 perform experiments with strains of different microorganisms. 1Newcastle University, Institute for Ageing and Health, Newcastle The quantitative determination of important parameters, such upon Tyne, United Kingdom; 2Newcastle University, School as metabolite concentrations in fermentation experiments is a of Mathematics and Statistics, Newcastle upon Tyne, United major task for the set up and validation of mathematical models. Kingdom; 3Newcastle University, School of Computing Sciences, Here, experimental methods for sampling, sample preparation Newcastle upon Tyne, United Kingdom and sample analysis are being developed. The functional units under investigation are mainly focused on global control and The Centre for Integrated Systems Biology of Ageing and Nutrition signal transduction systems. Examples in prokaryotes are: (CISBAN) investigates the network of mechanisms contributing catabolite repression and two component signal transduction in to cellular ageing in vitro, and how cell defects contribute to E. coli, redox control in Rhodospirillum rubrum and phototaxis ageing in vivo, as well as how these processes are modulated in Halobacteria. The TNF induced apoptosis in mammalian cells by nutrition. We use human cell cultures, ageing mice and the as well as cell cycle regulation in yeast and in hepatocytes are yeast Saccharomyces cerevisiae, as model organisms, and target examples of signal transduction units in eukaryotic cells under systems fundamental to the ageing process, including the roles investigation. of mitochondrial dysfunction, telomere erosion, oxidative stress Different modeling strategies are applied: Qualitative models and protein homeostasis. Our approach is highly interdisciplinary, are set up when kinetic information is scarce and time resolved with bioinformaticians, statisticians and computational modellers data is missing. Models analyzed so far focus on large metabolic

working closely with laboratory biologists in an iterative cycle of or signal transduction networks. Quantitative models are used Arenas data generation, analysis and modelling. when only small parts of the networks are analyzed and time CISBAN research is organised through three closely connected course data is available. Besides the set up of very detailed programmes: ‘Functional Genomics and Proteomics’, ‘Integration and comprehensive models, a further topic is the assessment of Partial Systems’ and ‘In silico Systems Biology’. These are of models and of the kinetic parameters that are obtained founded on a substantial core of basic and clinical research on after parameter identification. The modeling and simulation ageing and nutrition within the Institute for Ageing and Health environment ProMoT/Diva is developed and adapted for cellular in partnership with the Schools of Computing Science and systems. The tool allows an easy set up of models via predefined Mathematics & Statistics, and the UK’s North East Regional modeling objects and also allows a visualization of large networks. e-Science Centre (NEReSC). Homepage: http://www.mpi-magdeburg.mpg.de/research/ Functional Genomics and Proteomics: provides large amounts groups/sb Address: MPI Magdeburg, Sandtorstr. 1, 39106 of data using post-genomic technologies in a hypothesis- Magdeburg, Germany Contact: Dr. A. Kremling (kremling@mpi- independent fashion. Generates genomic and proteomic data on magdeburg.mpg.de) telomere-dependent senescence in yeast and human cells and in mice under ageing and dietary restriction. A-48 Integration of Partial Systems: uses a variety of functional assays to test the impact of cellular damage and damage response ERASysBio - A funding initiative to advance the European pathways for ageing at cellular, tissue and, ultimately, organism research area in their networking in the field of systems level. biology In silico Systems Biology: develops and applies new mathematical Heidelberger, Maike1; Simons, Veronika2 modelling, statistical inference and data management 1SysMO, , Germany; 2ERASysBio, Germany programmes. CISBAN is also home to the Biology of Ageing e-Science Integration and Simulation (BASIS) system providing The overall objective of the ERA-NET (European Research Area a flexible SBML-based platform for model development and Network) scheme is to step up the cooperation and coordination stochastic simulation, as well as to the developing SyMBA data of research activities carried out at national or regional level in portal for integration of experimental data. the EU Member States and associated countries. This is to be CISBAN is funded primarily by the UK Biotechnology and achieved through networking and by synchronising the different Biological Sciences Research Council (BBSRC). underlying national funding mechanisms. ERA-NETs are designed to enable the participating partner countries and national systems A-47 to engage in scientific tasks together that they would not have been able to deal with independently. Max-planck-institute of complex technical systems - ERASysBio is a consortium of 15 ministries and funding systems biology group organisations from EU member states and associated countries. Kremling, Andreas The partners of the consortium are aiming to coordinate their Max-Planck-Institute of Complex Technical Systems, Systems national research programmes in systems biology and agree on a Biology Group, Magdeburg, Germany common European research strategy. Currently, the first ERASysBio Call “Systems Biology of The Systems Biology Group of the MPI Magdeburg (head of Microorganisms (SysMO)” is the largest and most ambitious group E.D. Gilles) has a strong focus on combining modern Systems Biology funding initiatives in Europe. It was launched modeling approaches with sophisticated experimental methods in 2005 as the first step in developing a number of transnational for a better understanding of cellular systems. research activities. A subset of six funding organisations jointly Research in modern molecular biology is mainly descriptive and is funding eleven consortia, including 87 working groups across qualitative. To understand the complex structure and behavior of Europe, under this initiative with a financial commitment across metabolism, signal transduction and regulation requires an overall all partners of approximately €30.5M. The transnational SysMO holistic analysis. Due to the multitude of interacting components, consortia started in March 2007 and aim to understand the an understanding of these processes just by reflection is not dynamics of biological properties in different microbes with high possible. Here, the aid of mathematical modeling is absolutely relevance for biotechnology, health and nutrition. A central data essential to combine the available biological knowledge with a management, SysMO DB, has just started to enable data storage system-theoretical way of thinking. After starting research work in and sharing across the borders of consortia and participating Magdeburg in 1998, the Systems Biology group has continuously countries. extended its activities by taking up studies on a number of The ERASysBio consortium is planning two further transnational

ICSB 2008 57 calls for research projects in systems biology. the EU that involves nine research groups from Europe and the www.erasysbio.net www.sysmo.net US. Veronika Simons is the coordinator of the transnational funding initiative for systems biology - ERASysBio. She works for the A-51 German partner PtJ: project management agency for the German Federal Ministry of Education and Research. Representing evidence for interacting network elements Maike Heidelberger is coordinating the transnational call SysMO (REFINE) on behalf of the participating European funding organizations. She Ananiadou, Sophia1; Duncan, Hull2; Douglas, Kell2; Pedro, also works for the PtJ. Mendes2; Steve, Pettifer2; Jun’ichi, Tsujii3; Yoshimasa, Tsuruoka2; Alice, Villeger2 A-49 1University of Manchester, School of Computer Science, Manchester, United Kingdom; 2University of Manchester, VTT systems biology and bioinformatics - a framework for Manchester, United Kingdom; 3University of Tokyo, Tokyo, Japan the integration of biotechnology, information technology and engineering We present tools created by the REFINE project http://www. Oresic, Matej; Merja, Penttilä; Jyrki, Lötjönen; Kaisa, Poutanen; dbkgroup.org/refine , which is combining text-mining tools Hans, Söderlund provided by the National Centre for Text Mining http://www. VTT Technical Research Centre of Finland, Espoo, Finland nactem.ac.uk such as Kleio and FACTA with Utopia tools for visualising, managing and querying bibliographic and biochemical VTT Technical Research Centre of Finland is an impartial expert data (biological pathways encoded in SBML http://www.sbml. organisation. Systems biology at VTT aims to integrate a broad org. This novel combination of tools allows users to find primary

Arenas range of in-house expertise. The systems biology platform scientific evidence in the PubMed literature database that given combines bioinformatics with molecular profiling approaches metabolic pathways and their metabolites exist. and new measurement technologies, and benefits from in- house knowledge across information technology, biosensors, A-52 microbiology, plants, nutrition, and medical domains. We will highlight four projects from different biotechnology and The E-Cell project medical domains where the VTT’s systems biology plays an Takahashi, Koichi1; Sakurada, Takeshi2; Koizumi, Moriyoshi3; important role: Okada, Chihiro4; Matsuzaki, Yuri5; Tomita, Masaru2 (1) The Finnish Centre of Excellence in White Biotechnology - 1the Molecular Science Institute, Berkeley, United States; 2Keio Green Chemistry Research is a Centre of Excellence suppoirted University, Institute for Advanced Biosciences, Fujisawa, Japan; by the Academy of Finland. Its main aim is to develop efficient 3Keio University and Japan Science and Technology Agency microbes and processes for the production of chemicals from (JST), Fujisawa, Japan; 4Mitsubishi Space Software Co. Ltd., plant biomass sugars. Bioinformatics Department, Amagasaki, Japan; 5Tokyo Institute of (2) PredictAD (From patient data to personalized healthcare Technology, Department of Computer Science, Tokyo, Japan in Alzheimer’s disease) is an EU FP7-ICT (Virtual Physiological Human) funded research project aiming at developing a The E-Cell Project is an international research initiative aiming standardised and objective solution that would enable an to model and reconstruct biological phenomena in silico, and earlier diagnosis of Alzheimer’s disease, improved monitoring of developing necessary theoretical supports, technologies and treatment efficacy and enhanced cost-effectiveness of diagnostic software platforms to realize precise cellular and biochemical protocols. simulations at the molecular level. Some of the research foci of (3) AtheroRemo (European Collaborative Project on Inflammation the Project include: and Vascular Wall Remodelling in Atherosclerosis) is an EU • Modeling methodologies, formalisms and techniques, including FP7-Health funded research project aiming to identify novel technologies to predict, obtain or estimate parameters such as inflammatory mechanisms in vascular remodelling by combining reaction rates and concentrations of molecules in the cell. the exploration of human biobanks with animal models and • E-Cell System, a software platform for modeling, simulation and established cellular models. This new knowledge will be used to analysis of complex, heterogeneous and multi-scale systems develop new preventive, diagnostic and therapeutic strategies like the cell. against atherosclerotic cardiovascular disease. • Numerical simulation algorithms — stochastic/deterministic, (4) SYSDIET (Systems biology in controlled dietary interventions spatial/non-spatial. and cohort studies) is a Nordic Centre of Excellence funded by • High-performance computing. Nordforsk. SYSDIET aims at a coordinated effort for exploitation Various types of modeling and analysis applications (Circadian of nutrigenomics/systems biology tools in human randomized Rhythm, Liver cells, Bacterial chemotaxis, Human erythrocytes, controlled dietary interventions and animal and cell culture studies Myocardium cell, etc.) are currently ongoing using outcomes of in order to find out novel mechanisms by which Nordic foods and the E-Cell project. diets could be modified to promote health and prevent chronic The E-Cell Project is open to anyone who shares the view with us diseases. that development of cell simulation technology, and, even if such ultimate goal might not be within ten years of reach yet, solving A-50 various conceptual, computational and experimental problems that will continue to arise in the course of pursuing it, may have a ComplexDis - an EU project aiming to find markers for multitude of eminent scientific, medical and engineering impacts personalized medication on our society. Benson, Mikael We will be having some hands-on demonstrations of our latest Gothenburg University, Gothenburg, Sweden software products including E-Cell IDE (Integrated Development Environment). More project details are showing at our website Common diseases like allergy, obesity and cancer are complex. URL: http://e-cell.org. Each of these diseases is caused by altered interactions between multiple genes. These alterations may differ between different individuals although they appear to have the same disease. A clinical consequence is variable response to treatment. Ideally, physicians should be able to personalize medication based on measuring a few protein markers in for example blood or saliva. The aim of ComplexDis is to identify such proteins using high-throughput technology, high-performance computing and systems biology. ComplexDis is a three year project funded by

58 ICSB 2008 A-53 different levels of the system (lung, blood, muscle and whole-body bioenergetics) with a multidimensional assessment (physiological, Systems and synthetic biology at caltech genomics, proteomics and metabolomics) in three subsets of age Hucka, Michael matched subjects: healthy sedentary individuals, COPD patients California Institute of Technology, Engineering and Applied with and COPD without muscle wasting. Science, Pasadena, United States Main bioinformatics challenges are those related with data mining and multilevel data integration; development of SBML The California Institute of Technology offers a wealth of standards; deterministic modelling of oxygen transport-central opportunities in Systems and Synthetic Biology. metabolism and mitochondrial ROS generation; development of The Biological Network Modeling Center (BNMC, http:// tools for probabilistic modelling; and, integration of the simulation bnmc.caltech.edu) brings together biologists, bioengineers, environment into the web-based Biobridge portal. The latter will mathematicians, and computer scientists to develop and apply cover four aspects: state-of-the-art tools for the mechanistic modeling of molecular •Integrative analysis of biomedical data to address well defined networks of all kinds. Our long-term goal is to provide biologists use cases with the computational tools they need to define families of • Prepared to serve several user-profiles models at any level of detail, efficiently analyze their properties, • Access to tools for inference analysis and simulations, and and relate them back to experimental data to help answer • Become a knowledge management tool interoperable with the question, ‘What is the next experiment that would best other ICT platforms differentiate between the current alternative hypotheses?” The bioinformatics platform must allow generalization, so that The Center for Biological Circuit Design (CBCD, http://cbcd. the simulation environment generated within the project is useful caltech.edu) develops new ways of designing, building and to address other complex clinical problems. In this regard, analyzing biological circuits and the control of information the project should pave the way for future personalized health flow within them. Our interdisciplinary group of biologists and strategies. engineers work together to 1) deduce simple rules of biological Supported by Biobridge (FP6-2005-LIFESCIHEALTH-037909)

circuits and understand their operation at the levels of molecules, Arenas cells, organisms and ecosystems, and 2) learn how to model, A-55 design, build and analyze biological circuits. We combine the experimental biologist’s desire to find key principles from the Expression time course analysis of the malaria parasite diversity of biological circuits, with the physicist’s sense of plasmodium falciparum measurement and of simple underlying mechanisms, and the Scholz, Matthias; Fraunholz, Martin engineer’s aesthetic of “to build is to understand.” University of Greifswald, Competence Center for Functional The Synthetic and Systems Biology (SSB, http://be.caltech. Genomics, Greifswald, Germany edu/academics/ssb.html) group in Bioengineering develops advanced techniques for analyzing, designing, and synthesizing Objective: The malaria parasite Plasmodium falciparum behavior within biological systems, with applications in biology, replicates asexually in an infection cycle within human biomedicine, and biotechnology. We emphasize design and erythrocytes (red blood cells). By using available time course analysis of circuit- and systems-level behavior, from molecular microarray data we can show that the gene expression data of switches to information processing networks, using techniques this intraerythrocytic developmental cycle (IDC) form a circular from a range of engineering disciplines including control theory, structure. The trajectory of the data is curved (nonlinear) and chemical engineering, computer science, physics, and applied closed (circular). To describe the trajectory by a closed curve, mathematics. we use circular principal component analysis which is a special Degree programs for PhD students include Biology, Biochemistry type of nonlinear PCA (NLPCA). Both methods are nonlinear and Molecular Biophysics, Chemical Engineering, and the extensions of standard (linear) PCA, based on auto-associative Systems and Synthetic Biology thrust of Bioengineering. neural networks. Results: Circular PCA provides a component which describes A-54 the principal curvature of the data by a closed curve. This circular component can then be used as a mathematical model BIOBRIDGE - Integrative genomics and chronic disease of the developmental cycle of the malaria parasite which gives phenotypes: Modelling and simulation tools for clinicians a continuous and noise-reduced description of the biological Roca, Josep1; Falciani, Francesco2; Maier, Dieter3; Selivanov, process. Vitaly4; Kalko, Susana4; Villà, Jordi5; Brozek, John6; Brugard, Jan7; Conclusions: The model was used to visualize the transcriptional Barreiro, Esther5; Turan, Nil8; Cascante, Marta4 activity on the parasite’s nuclear chromosomes. The analysis 1IDIBAPS - University of Barcelona, Barcelona, Spain; 2University revealed a distinct transcriptional activity of telomeric from of Birmingham, Birmingham, United Kingdom; 3Biomax, centromeric genes. This activity delay between genes of Munchen, Germany; 4IDIBAPS-University of Barcelona, chromosome ends (telomeres) and central chromosomal Barcelona, Spain; 5Univ. Pompeu Fabra - IMIM, Barcelona, Spain; regions suggests that key events of the IDC are initiated at the 6Genfit, Lille, France;7 Mathcore, Linköping, Sweden; 8University subtelomeric regions of the P. falciparum chromosomes. of Birmingham, Birmingham, Spain

Biobridge explores the underlying mechanisms of systemic effects in three highly prevalent chronic conditions: chronic obstructive pulmonary disease (COPD), congestive heart failure (CHF) and diabetes. Hallmark features of the systemic effects are skeletal muscle dysfunction and muscle wasting. It is hypothesized that the nitroso-redox unbalance of the cardiovascular system is a common mechanism in these three chronic disorders accounting for functional heterogeneities

(O2 uptake/O2 flow mismatching) at tissue level that would explain both skeletal muscle dysfunction (mechanical-energetic uncoupling) and abnormal tissue remodelling (muscle wasting) seen in phenotypes of these diseases associated with poor prognosis. The complexities of the biomedical problem, however, prompt the need for introducing novel strategies based on “in- silico” modelling. The experimental studies explore training-induced effects at

ICSB 2008 59

Dedicated Posters Posters Dedicated Dedicated session 1-1: and shaking flasks were done as 5 biological replicates. DI-MS Cell-regulation - metabolism technique and data analysis methods together with correlation analysis were used to find out the metabolites that significantly correlate in the response over time in the different C-sources DS1-1-09 which are glucose, glycerol, and ethanol. Both positive and negative ion mode DI-MS were used to obtain complete Toward a systems neurochemistry of brain synaptic information. The significant metabolites are chosen as the ones function carrying group information which is able segregate one or more Maciejewski, Paul groups at some level of significance. For all samples, based on Yale University, New Haven, United States the P-value calculated for each pair of ions, we got a symmetric matrix of correlations with the ions along both axes. We Objective: This project introduces elements of a systems investigated the informational content by Principal Component neurochemistry of brain synaptic function. Its immediate objective Analysis (PCA) to find out the metabolites that significantly is to represent well-defined systems of compartmentalized correlate in the response over time in the different C-sources. biochemical processes whose coordinate activities in principle There was a highly significant correlation between the groups of accomplish a specific, well-defined function — the sustained, significant metabolites i.e. amino acids, cyclic AMP, cyclic dAMP, unidirectional flux of a chemical neurotransmitter from presynaptic glycerol, and myo-inositol. The result from PCA shows that that nerve terminal to synaptic cleft to astrocyte. the main variance in the data is caused by the different C-sources Results: Present understanding of brain synaptic trafficking with really high reproducible. of chemical neurotransmitters focuses on a few key enzymes associated with neurotransmitter synthesis and degradation. DS1-1-11 Although these key enzymes are critical to understanding brain synaptic function, they provide an incomplete account of the Towards a detailed and quantitative mathematical biochemical basis of chemical neurotransmission. Computational description of the glycolytic pathway in Lactococcus lactis methods were devised to construct alternative biochemical Maclear, Athlee S.; Snoep, Jacky L.; Rohwer, Johann M. systems for synaptic trafficking of chemical neurotransmitters Stellenbosch University, Biochemistry, Stellenbosch, South Africa Dedicated consistent with stoichiometric principles and existent knowledge Posters about enzyme and carrier protein expression, localization and Objective: Lactococcus lactis is the preeminent model organism function. These methods were used to construct biochemical within the lactic acid bacteria — a group of microorganisms of systems for brain synaptic trafficking of glutamate and GABA, prime industrial significance, finding considerable and wide- major excitatory and inhibitory neurotransmitters, respectively, of ranging application in the manufacture of fermented food the human brain. Results indicate that key enzymes associated products, where the resultant acidification contributes to the with glutamatergic and GABAergic neurotransmission do not preservation, texture and organoleptic quality of the food, and function in isolation, but rather in the context of specific, well- increasingly in other production processes. Thus, from an defined, distributed systems of enzymes and transporters. industrial and biotechnological stand point there is great interest Alternative biochemical systems for brain synaptic trafficking in optimizing the production of the varied end-products of L. of glutamate and GABA may represent adaptive responses to lactis energy metabolism. Despite some success and extensive environmental, genetic, and pharmacological perturbations. experimentation in this regard, a thorough understanding of the Conclusions: This work provides a sound conceptual, functioning of L. lactis at a systems level is still lacking – it remains empirically-based foundation for future theoretical, computational, unclear where glycolytic flux control resides and what governs the and experimental investigations of brain synaptic function from a shift from homolactic to mixed acid fermentation in this organism. biochemical systems perspective. Although the work focuses on Results: Progress to date in this ongoing investigation includes:

a specific metabolic function (synaptic trafficking of a chemical the experimental determination of Vmax parameter values for all neurotransmitter) of a specific organ (the brain), its methods are the glycolytic enzymes of the wild-type strain, L. lactis MG1363, quite general, and readily applicable to investigations of metabolic and their inclusion into the existing glycolytic kinetic model; the functions of interest in other biological systems. calculation of a new steady state using PySCeS and comparison of the resultant fluxes to their experimentally-determined DS1-1-10 counterparts, as reported in literature; and fitting of these parameter values, where necessary. Additionally, generalised Dynamic metabolic footprinting in yeast Saccharomyces supply and demand analysis was performed on the updated cerevisiae model. Preliminary results seem to support deductions made from Chumnanpuen, Pramote1; Hansen, Michael Adsetts Edbe2; the reported experimental data, specifically with respect to the Smedsgaard, Jorn3; Nielsen, Jens1 control exerted by the glycolytic enzymes, lactate dehydrogenase, 1Chalmers University of Technology, Department of Chemical pyruvate kinase and glyceraldehyde-phosphate dehydrogenase. and Biological Engineering, Gothenburg, Sweden; 2Technical Conclusion: It is hoped that this work will contribute towards the University of Denmark, Department for Systems Biology, Lyngby, realisation of a holistic understanding of energy metabolism in L. Denmark; 3Technical University of Denmark, National Food lactis — crucial to the quantitative and accurate anticipation of Institute, Lyngby, Denmark how the organism will respond to changes in its environment, and thus essential for any truly successful rational intervention strategy Since the word metabolism comes from Greek metabolé which to reroute and/or optimize carbon fluxes. means change or transformation, the levels of many metabolites change with half times of minutes, seconds or even faster. The DS1-1-12 complexity and dynamics nature of metabolite are increasingly apparent. Functional metabolic modularity of amine metabolism Metabolomics is rapidly becoming one of the cornerstones of Reyes-Palomares, Armando; Montañez, Raúl; Abrighach, Hicham; functional genomics and systems biology. Metabolite profiling Urdiales, José Luis; Sánchez-Jiménez, Francisca; Medina, Miguel has profound applications in discovering the mode of action Angel of drugs or herbicides, and in unravelling the effect of altered ProCel Lab. F. Sciences. University of Málaga and CIBER de gene expression on metabolism and organism performance in Enfermedades Raras (CIBER-ER), Department of Molecular biotechnological applications. Recently, Direct Infusion - Mass Biology and Biochemistry, Málaga, Spain Spectrometry (DI-MS) is going to be an alternative to GC-MS and seems to be an ideal analytical tool for high-throughput Objective: Our group has been working for the last 20 years on metabolome analysis. Here we studied the dynamic metabolic different aspects of metabolic pathways related to amino acids footprinting study in yeast S. cerevesiae strain BY4709 (wild and to their amine derivatives in mammals. These pathways type) at 10 time points, cultured in different 3 types of C-sources, are in the origin or are responsible of the symptoms of many

62 ICSB 2008 diseases. Nitrogen metabolism involves hundreds of enzymes DS1-1-14 and transporters (some of them poorly characterized), frequently expressed following a tissue-specific pattern. Therefore, nitrogen The regulatory function of metabolic channeling: ethanol metabolism is a Complex System requiring systemic approaches cycle in Zymomonas mobilis to be understood as a whole. Since metabolic networks Kalnenieks, Uldis; Galinina, Nina; Rutkis, Reinis; Toma, Malda are modular, nitrogen metabolism network could be built by Maija integration of discrete modules or pathways. This is our objective University of Latvia, Institute of Microbiology and Biotechnology, for the work described in this communication. Riga, Latvia Results: Mathematical modelling through systems of ordinary differential equations (ODEs) was carried out in Perl programming Objective: The regulatory role of the two alcohol dehydrogenase language. Once each individual model was validated by contrast (ADH) isoenzymes in the facultatively anaerobic ethanol-producing with experimental data, it was incorporated into a general bacterium Zymomonas mobilis was studied. Perturbation of the integrated system and migrated to SBML. At the moment, we steady-state in aerobic chemostat culture of this bacterium with have successfully integrated mammalian polyamine and histamine a small amount of ethanol has been shown to cause a burst metabolism, arginine catabolism and methyl and folate cycles. of respiration, although the reactant ratio of the ADH reaction: Simulations of our model suggest a relevant role of S-adenosyl ([NADH][acetaldehyde][H+])/([ethanol][NAD+]) remains well above methionine (SAM) and acetyl-CoA availability in polyamine the equilibrium constant during the experiment, hence favoring homeostasis, which are not usually considered in systemic ethanol synthesis [1]. To explain this finding, a novel redox experimental studies. Furthermore, competence for SAM seems futile cycle, named “ethanol cycle”, has been suggested in this to contribute to the previously observed antagonism between bacterium, performing simultaneous catalysis of ethanol synthesis polyamine and histamine metabolism. and oxidation by the two ADH isoenzymes, ADH I (adhA) and Conclusions: This communication shows that metabolic ADH II (adhB). Both reactions proceed several times faster than modularity in the context of mammalian amine metabolism is fully the net ethanol synthesis. We have proposed that ADH I is operative. This work illustrates how Systems Biology (applied to a catalyzing ethanol synthesis, while ADH II (or, at least a part of metabolic modelling framework) helps to suggest new hypothesis the isoenzyme molecules) in a more oxidizing microenvironment that need to be experimentally validated to add new information catalyzes oxidation of ethanol and channels NADH to the to complex biological systems. respiratory NAD(P)H dehydrogenase. This was confirmed by construction of an ADH II-deficient adhB::kan( r) mutant [2], and [Supported by Spanish MEC (SAF2005-1812), Fundación Ramón Areces and “Plan Andaluz showing that it had lost the respiratory response to the ethanol de Investigación” (BIO-267, P07-CVI-02999 and CVI-657). The CIBER de Enfermedades Raras is an initiative of the ISCIII.]. pulse.

Results: In the present work we show that ADH II contributes Posters DS1-1-13 to the regulation and buffering of the intracellular reduced Dedicated nicotinamide dinucleotide pools: the ADH II-deficient mutant Dynamics on the cell membrane; consequences of protein was unable to maintain homeostasis of the intracellular NADH interactions and NADPH pools during transition from anaerobic to aerobic Guseva, Ksenia1; Thiel, Marco2; Grebogi, Celso3 conditions. 1University of Aberdeen, Physics, Aberdeen, United Kingdom; Conclusions: We conclude that ethanol cycle represents a 2University of Aberdeen, Aberdeen, United Kingdom; 3University novel regulatory mechanism for respiratory metabolism, aimed of Aberdeen, Aberdeen, United Kingdom at buffering of the pools of NADH and NADPH. This type of mechanism might be particularly efficient for the purposes of Objective: The cell membrane in procaryotic organisms plays rapid, short-term regulation, before the tuning of metabolic fluxes a crucial role in the cell’s metabolism,and it is known to have at the transcriptional level takes place. complex molecular organization. The commonly accepted mosaic [1] U. Kalnenieks et al. (2002) FEBS Lett 522, 6-8. [2] U. fluid model has recently been undergoing a process of revision. Kalnenieks et al. (2006) FEBS Lett 580, 5084-5088. These recent studies are focused on the fact that proteins do not diffuse freely in the membrane, but form clusters as they DS1-1-15 mutually interact. The interactions can be of different types. Relevant forces have enthalpic (elastic) and entropic (fluctuation Functional state analysis of the genome-scale Escherichia induced - Casimir-like) origins, and experimental evidence about coli regulatory network the influence of the membrane curvature suggests that these Gianchandani, Erwin; Papin, Jason effects can have important consequences for protein motion. The University of Virginia, Biomedical Engineering, Charlottesville (VA), aim of our work was to construct a model that elucidates how the United States dynamics of proteins in the membrane, can change properties such as gating thresholds for ion channels, due to proteins Objective: A transcriptional regulatory network constitutes the interactions. collection of regulatory rules that link environmental cues to the Results: Our work involves computer simulations of randomly transcription state of a cell. We previously proposed a matrix moving particles with time depending, attractive and repulsive formalism that quantitatively represents these rules and allows forces acting on them. We study the movement of the interacting for systemic characterization of properties of any transcriptional particles in spaces of different topologies and observe the regulatory system (TRS) [1]. In particular, the formalism facilitates formation of aggregated clusters for typical magnitudes and types the computation of the transcription state of a genome under any of forces between them. We obtain qualitative and quantitative set of environmental conditions. To demonstrate the applicability information about how the spatio-temporal movement changes of this method to genome-scale systems, we have applied it to the gating properties of the channels. the genome-scale TRS of the model prokaryote Escherichia coli. Conclusions: In agreement with experimental results, our Results: Our genome-scale reconstruction of E. coli model shows that the interaction between inclusions in the cell transcriptional regulation was comprised of 126 environmental membrane influences the gating mechanism of ion channels stimuli such as oxygen and glucose, affecting 612 target genes. substantially. Furthermore, our model suggests that the forces of These target genes in turn influence 1919 metabolic enzymes and the interaction between the inclusions make it necessary to refine transporters. We generated the regulatory network matrix R that the mosaic fluid model. We describe the protein dynamics on the captures these experimentally-characterized rules and analyzed cell’s surface, and the aggregation of proteins in the membrane in the fundamental subspaces of the matrix (i.e., the null, left null, accordance to recent experimental results. We also argue that the column, and row spaces) to describe properties of the system. complex dynamics of membrane proteins will effectively change For example, the null space of R captured all possible balanced chemical reaction constants. We furthermore conclude that the expression states of the E. coli TRS, while the left null space of gating of mechanosensitive channels, which is crucial for the cell’s R contained groups of genes that are coordinately regulated and osmoregulation, is altered by these processes. can be classified as “regulated units” or regulons. We further

ICSB 2008 63 used Monte Carlo sampling to characterize the range of variation network is shown on different levels of granularity. The new in E. coli transcription states across possible environments, system of outlines integrates the most useful features of many and validated some of these in silico transcriptional profiles by different approaches and has at least three main functions: 1) comparison with available expression data. navigation through the constantly updated network; 2) intelligent Conclusions: Ultimately, analysis of the E. coli TRS with the overview and 3) each outline unites lower-level diagrams into one regulatory network matrix formalism resulted in novel findings large virtual map. about the system that may be further experimentally validated, and demonstrated how the regulatory network matrix formalism DS1-1-18 extends to genome-scale systems. Reference: 1. Gianchandani EP et al. 2006. PLoS Comput Biol Localization studies of Ena1p provide mechanistic insights 2: e101. into the salt sensitivity of vacuolar protein sorting mutants Logg, Katarina1; Warringer, Jonas2; Hashemi, Sayed Hossein2; DS1-1-16 Kall, Mikael1; Blomberg, Anders2 1Applied Physics, Göteborg, Sweden; 2Cell and Molecular Biology, Comparative systems biology: The differences can be in Göteborg, Sweden the interactions, not the components Teusink, Bas1; Hugenholtz, Jeroen2; Westerhoff, Hans3 Objective: In a phenotypic screen on budding yeast 1Netherlands Institute for Systems Biology, IBIVU - Free University (Saccharomyces cerevisiae) an unexpectedly large part of the Amsterdam, Amsterdam, Netherlands; 2Netherlands Institute deletion mutants that scored for salt sensitivity, were found to Systems Biology, Faculty of Science - University of Amsterdam, be vacuolar protein sorters (vps). The main component in the Amsterdam, Netherlands; 3Netherlands Institute Systems Biology, sodium homeostasis system is the sodium exporter Ena1p and Mol. Cell. Physiol. - Free University Amsterdam, Amsterdam, therefore we hypothesized that the vps proteins are involved in Netherlands the localization of Ena1p to the plasma membrane. By means of fluorescence microscopy, GFP-tagged Ena1p was monitored Objective: Comparative analyses, as demonstrated by for selected vps mutants and the localization, quantity and comparative genomics and bioinformatics, are extremely powerful transportation were analyzed for comparison with the wild type. Dedicated (i) for transfer of information from well-studied organisms to Results: Overexpression of Ena1p did suppress the salt Posters the other organisms, and (ii), when coupled to functional and sensitivity for some of the deletion mutants, which indicates phenotypic information, for insight in the relative importance that there indeed is a connection between Ena1p and the salt of components to the observed differences and similarities. sensitivity of the vps deletion mutants. In the wild type Ena1p Within the SysMo project “Comparative systems biology of was localized to the plasma membrane as well as to the Lactic Acid Bacteria” we hypothesize that important aspects of vacuole, probably for breakdown. For all mutants, Ena1p was the functional differences between organisms derive not only correctly localized to the plasma membrane but the localization from the differences in genetic components (which underlies to the vacuole varied between the mutants. A correlation was comparative genomics) but also from the interactions between found between the salt sensitivity and the kinetics of Ena1p their components. Therefore, this SysMo project will develop localization to the plasma membrane after ENA-GFP plasmid Comparative Systems Biology by comparing the primary onset. Quantitative differences in plasma membrane GFP signal metabolism of three related, but functionally different, lactic acid were also measured and the most salt sensitive deletion strain bacteria, i.e., Lactococcus lactis, Streptococcus pyogenes and tested showed a substantially increased GFP signal in the plasma Enterococcus faecalis. To illustrate the power of comparative membrane. systems biology in general, we here compared only the regulation Conclusions: Our data is not conclusive, but provide clues of the first part of glycolysis, but between many different about mechanisms behind the salt sensitivity of the vps deletion organisms, from bacteria to yeast to man. mutants. One possible explanation is the temporal mislocalization Results: The type of feedback onto either the glucose transporter of Ena1p to the plasma membrane. Another explanation is that itself (PTS system in prokaryotes), or the subsequent hexokinase the activation of Ena1p in the plasma membrane is affected as step (in case of glucose transport by facilitated diffusion, yeast a result of changes in plasma membrane lipid composition in and human), or the absence of feedback (trypanosomes, the vps mutants. Signalling defects in downstream components pancreatic beta-cells), could be related to the autocatalytic design involved in salt tolerance is also a suggested explanation. of glycolysis, the environmental conditions, and the functional role of glycolysis in these different organisms or cell types. DS1-1-19 Conclusions: Comparative systems biology recognizes that differences in phenotype can arise not only from differences in the Relationship between adenine nucleotide pool size and parts lists, but also from the interactions between these parts. glycolytic flux in Saccharomyces cerevisiae Comparative systems biology will be useful not only for deeper Westoll, Julian; Pillay, Che; Snoep, Jacky; Rohwer, Johann understanding of design principles through comparison, but also Triple-J Group for Molecular Cell Physiology, Biochemistry for understanding the behavior of different organisms and cell Department, Stellenbosch University, Stellenbosch, South Africa types to drugs and disease. Objectives: Glycolysis is one of the most well-characterised DS1-1-17 metabolic pathways in the cell. Detailed information is available on the mechanisms and regulation of all of the glycolytic enzymes. The system of outlines for human metabolic network In spite of this we do not yet have a complete picture of how Mazein, Alexander; Goryanin, Igor the carbon flux through glycolysis is controlled. It has long The University of Edinburgh, Edinburgh, United Kingdom been known that the energy charge (as reflected by the relative concentrations and distribution of ATP, ADP and AMP) plays a role Systems for navigation through metabolic network are usually in glycolytic flux control. In this study we investigated whether the either a list of pathways that are subdivided into several groups total adenine nucleotide pool size (as opposed to ratios between or very simple outlines that generally repeat the list of pathways. specific adenylates) has a regulatory role. A kinetic model of yeast On the other hand, most useful attempts to overview a metabolic glycolysis was used to determine whether potential changes in network are either a very large map that fails to represent all the pool size would have an effect on the flux through glycolysis. known metabolic events any way, or a small “textbook” diagram An experimental strategy was devised to quantify changes in that underline only the most important relations. We introduce adenylate pool size in response to environmental perturbations. a new system of outlines that has been developed on the basis Results: The model parameter specifying adenylate pool size of the Edinburgh Human Metabolic Network that consists of was scanned from 2.5 to 13 mmol.L-cytosol-1. Glycolytic flux was more than 3000 reactions. Outlines and detailed diagrams are found to have a maximum value of 138 mmol.min-1.L-cytosol-1 at organized into a single system, where the human metabolic a pool size of 3.14 mmol.L-cytosol-1. As the adenylate pool size

64 ICSB 2008 was increased this value decreased to a minimum value of 77.63 the cra modulon, hierarchically regulates eleven operons encoding mmol.min-1.L-cytosol-1. Control analysis revealed that this effect genes of the glycolysis, TCA cycle and glyoxylate shunt. The was mediated mainly by changes in the activity of hexokinase. transcription of these genes was modeled, taking into account the An experimental extraction and assay method for adenine binding reaction of the Cra regulator protein and its DNA-binding nucleotides from yeast cells was developed and optimised. The sites as well as its inhibition by fructose 1,6-bis(phosphate). final protocol involved perchloric acid extraction followed by The intracellular fructose 1,6-bis(phosphate) concentration is bioluminescent assay with luciferase. demonstrated to decrease strongly during the fed-batch process. Conclusion: A kinetic model of yeast glycolysis predicts that The unknown kinetic parameters of the DNA-binding reaction for glycolytic flux is sensitive to changes in the total adenylate all genes of the network are determined by a statistical method, pool size. Ongoing work on this project involves experimental which predicts the parameters using DNA sequence data. The validation of this prediction, as well as evaluating its implications approach is introduced and the model simulation is illustrated. for the regulation of the adenylate pool size under different growth The simulation results are compared with the intracellular mRNA conditions. concentrations during fed-batch cultivation that were obtained by quantitative PCR analysis. DS1-1-20 Conclusions: A reasonable estimate of the kinetic parameters could be obtained using the presented approach. The modeling Bakers yeast as a prototype for metabolic syndrome concept can be applied to hierarchically regulated networks and Vemuri, Goutham1; Nielsen, Michael N.2; Hellgren, Lars I.2; provides the necessary interfaces for linking the regulatory with Nielsen, Jens1 the metabolic networks. 1Chalmers University of Technology, Department of Chemical and (1) Hardiman et al. J Biotechnol 132 (2007) 359-74. Biological Engineering, Göteborg, Sweden; 2Denmark Technical University, Department of Systems Biology, Lyngby, Denmark DS1-1-22

Several proteins in the bakers yeast, Saccharomyces cerevisiae, Multi-level modelling of the parallel metabolism of ethanol have counterparts in humans with conserved function and and retinol, with implications for foetal alcohol syndrome regulation. The metabolic pathways in which these proteins Hellgren, Mikko1; Nordling, Torbjorn E.M.2; Jacobsen, Elling W.2; participate are usually essential for a continued robust Hoog, Jan-Olov1 performance by the cells. In humans, a malfunction in the 1Karolinska Institutet, Dept. of Medical Biochemistry and regulation of these pathways triggers metabolic imbalance and Biophysics, Stockholm, Sweden; 2Royal Institute of Technology, eventually leads to disease. For practical and ethical reasons, it Automatic Control Lab, Stockholm, Sweden is not feasible to study the disease mechanisms in humans and Posters therefore, model organisms such as mice, fruit flies and even Objective: Models of the human metabolism are important for Dedicated bakers yeast, are employed to understand these processes. understanding diseases and could serve as a powerful tool in Although, multicellular model organisms offer the advantage of the drug discovery process. The complexity of even a unicellular evolutionary proximity to humans, it is not possible to rapidly organism is tremendous and most researchers have therefore quantify the dynamic phenomena of molecular interactions. We limited their modelling efforts to bacteria, or single intracellular studied the impact of dynamically changing nutrition environments pathways. We studied the parallel metabolism of ethanol and on the cellular energy level and how this controls lipid metabolism retinol in humans, because of its suggested physiological in S. cerevisiae. Our focus was on the regulation of cellular importance for the development of foetal alcohol syndrome. Large energy as well as the rate-limiting enzyme in lipid synthesis, acetyl ethanol intake will inhibit the conversion of retinol into retinoic CoA carboxylase, in response to the availability of nutrients that acid, which is a crucial transcription factor during embryonic mimic common human dietary scenarios. Our results indicate development. In this study the objective was to construct a that despite being a unicellular organism, the regulation of lipid quantitative model that connects phenotype observations at synthesis by cellular energy is highly conserved compared to a population, organic and intracellular level with differences in human, and that the yeast system makes a useful model for genotype and ethanol metabolism, for further prediction of the studying the molecular mechanisms of the metabolic syndrome influence on the foetus. as well as for screening new drug candidates and discovering Results: We constructed a multiple compartments model, new drug targets. which included a detailed description of the ethanol and retinol metabolism in hepatic cells for different genotypes. The model DS1-1-21 has been validated using published time-series measurements of ethanol, acetaldehyde and acetate concentrations in the Hierarchical regulation of the central carbon metabolism blood. This model correctly accounts for differences in geno- and of E. coli: Regulatory network structure and dynamic phenotype observed within the human population. Furthermore, modeling the model shows that the retinol metabolism is decreased Hardiman, Timo; Lemuth, Karin; Siemann-Herzberg, Martin; by ethanol ingestion, both via a reduced intracellular NAD+ Reuss, Matthias concentration, and by an inhibition of alcohol and aldehyde University of Stuttgart, Institute of Biochemical Engineering, dehydrogenases. Stuttgart, Germany Conclusions: We considered the problem of multi-level modelling with a human model for the ethanol and retinol Objective: E. coli fed-batch processes are often applied for metabolism in different compartments. This links intracellular production of recombinant proteins and bacterial metabolites. mechanisms to macroscopic observations. The model explained As in many other biotechnical processes this process operation the connection between geno- and phenotype differences leads to serious carbon limitation of growth. This in turn induces observed at a population level. This model also shows a plausible several well known regulation phenomena resulting in a decrease relationship between ethanol and retinol metabolism for e.g. foetal of biomass yield, specific productivity and cell viability. While alcohol syndrome. changes in the central carbon metabolism are very import for production processes, their experimental quantification is fragmentary and the mathematical description of the underlying regulatory processes are lacking. Results: Previously, we have discussed the structure of the regulatory network with the aid of time series analysis of the changes in the metabolic flux distribution and in the genome-wide transcript levels in fed-batch cultivations of E. coli K-12 W3110 applying a constant feed rate (1). At least three global regulatory systems are important for modeling the regulation. One of these,

ICSB 2008 65 DS1-1-23 DS1-1-25

Human metabolic network reconstruction for drug Have metabolic networks maximized molar yield during discovery evolution? Ma, Hongwu; Goryanin, Igor Schuster, Stefan1; Pfeiffer, Thomas2; Fell, David A.3 University of Edinburgh, Edinburgh, United Kingdom 1University of Jena, Dept. of Bioinformatics, Jena, Germany; 2Harvard University, Dept. of Organismic and Evolutionary Biology, Objective: Many diseases are caused by the deficiency of Cambridge, United States; 3Oxford Brookes University, School of metabolic enzymes and the subsequent accumulation of toxic Biological and Molecular Sciences, Oxford, United Kingdom substances or the lack of essential metabolites. A large number of metabolic enzymes have been selected as drug targets for Objective: Network analysis of metabolic systems allows one disease treatment. Therefore the study of human metabolism is to calculate possible metabolic flux distributions in the absence of great importance in researching the disease mechanism and of kinetic data. In order to predict which of the possible fluxes finding effective treatments. Specifically a high quality genome are present under given conditions, additional constraints and scale human metabolic network is desirable for studying the optimization principles can be used. One approach to computing diseases from a system level. unknown fluxes (frequently called Flux Balance Analysis) is Results: We reconstructed a high quality human metabolic based on the optimality criterion of maximizing the molar yield of network by integrating genome annotation information from biotransformations (even though this is not always clearly stated), different databases and metabolic reaction information from usually the ATP or biomass yields. literature. The network contains 2823 reactions, 2671 metabolites Results: The relevance and applicability of that approach are and 2322 genes which are organized into about 70 human critically examined, and it is compared with the principle of specific metabolic pathways. We found that more than 30% of the maximizing pathway flux. Diverse experimental evidence indicates genes in the network are related with one or more diseases in the that, often, those pathways are operative that allow fast but low- OMIM and GAD (Gene Association Database) databases. These yield synthesis of important products, such as fermentation in S. genes account for more than half of the reactions in the network. cerevisiae and several other yeast species. The dimorphic fungus, Nearly 400 of the proteins in the network have been selected Mucor racemosus mainly uses fermentation in the unicellular Dedicated as drug targets. The most targeted pathways include purine stage and respiration in the multicellular stage. Lactobacillus Posters metabolism, tyrosine metabolism and tryptophan metabolism, plantarum mainly relies on homolactic fermentation, although the glutamate-derived amino acid metabolism, bile acid synthesis, pathway of acetate production present in that bacterium exhibits androgen and estrogen biosynthesis, etc. Many of them are a higher yield. Another line of reasoning is based on evolutionary targeted to control the concentration of signal metabolites which game theory. It shows that flux maximization is often more can bind to protein receptors and through signal transduction important than maximization of molar yield. Therefore, organisms pathways affecting the expression of many genes. can be trapped in a Tragedy of the Commons, in which resources Conclusion: A high human metabolic network is reconstructed are not utilized economically. and provides a new platform to study the disease mechanism Conclusions: We conclude that the optimality assumption at a system level, thus paving the way to systems oriented drug underlying FBA is justified in many, yet by no means in all discovery. situations. Thus, maximization of molar yield is not a universal principle. DS1-1-24 References T. Pfeiffer, S. Schuster: Game-theoretical approaches to studying the evolution of biochemical systems. A new elementary flux based approach to Flux Balance Trends Biochem. Sci. 30 (2005) 20-25. Analysis S. Schuster, T. Pfeiffer, D.A. Fell: Is maximization of molar yield in Bordel Velasco, Sergio; Nielsen, Jens metabolic networks favoured by evolution? J. theor. Biol., epub Chalmers University, Göteborg, Sweden ahead of print (2007). S. Schuster, D.A. Fell: Modelling and simulating metabolic A new approach to the characterization of metabolic states networks. In: Bioinformatics: From Genomes to Therapies (T. using FBA is proposed. FBA is based on the optimization of an Lengauer, ed.), Vol. 2, Wiley-VCH, Weinheim 2007, pp. 755-805. objective function (normally growth rate) subject to a series of equality constraints (mass balances) and inequality constraints DS1-1-26 (the limits of some reaction rates). In many cases only a small number of inequality constrains are known experimentally, usually Large-scale expression screening reconstructs the heme those related to substrate uptakes. biosynthesis pathway Elementary flux modes are minimal sets of reactions satisfying the Nilsson, Roland1; Schultz, Iman2; Pierce, Eric2; Paw, Barry2; mass balances at steady state. Working in the elementary flux Mootha, Vamsi1 mode space allows considering only the inequality constraints 1Massachusetts General Hospital, Center for Human Genetics, and then simplifying the optimization problem. This representation Boston, United States; 2Brigham & Women’s Hospital, shows that the number of active elementary flux modes Hematology Division, Boston, United States characterizing a metabolic state is equal or lower than the number of inequality constraints imposed to the model. Therefore, the Objective: A central task for systems biology is to uncover traditional FBA approach, when used only with one constrained missing components of biological pathways. Biosynthesis of uptake flux gives as a result a single elementary mode. the oxygen-binding molecule heme is crucial to red blood cell The natural objective function to be maximized in any non- physiology, and many important enzymes such as those in the equilibrium chemical system is the dissipation function. Therefore mitochondrial electron transport chain rely on heme. Defects any other apparent objective function is the result of regulation in heme biosynthesis cause anemic diseases and porphyrias. achieved by modifying inequality constrains. An algorithm capable While the enzymes in this pathway have been thoroughly to predict a metabolic state from a set of inequality constraints studied, several transporters involved in the trafficking of has been proposed. intermediate metabolites across the mitochondrial membranes This approach is used to elucidate the regulatory mechanisms remain unknown. Because heme biosynthesis is strongly under the aerobic-anaerobic switch in Saccharomyces cerevisiae. transcriptionally regulated, we reasoned that it should be possible to discover these missing components from microarray data sets interrogating suitable experimental contexts. Results: We have developed a robust, context-specific method for large-scale screening of thousands of human and mouse microarray data sets using a form of Bayesian data integration. When applied to the heme biosynthetic pathway and combined

66 ICSB 2008 with high-confidence predictions of mitochondrial localization, Saccharomyces cerevisiae as a model to study the mechanisms this method identified three recently described iron and porphyrin of cellular aging in multicellular eukaryotes. Our objective is to transporters and predicted an additional three novel transporters. implement our understanding of yeast aging for the identification In addition, the method predicted two candidates likely to be of chemical compounds that can extend the life span of yeast involved in regulation of heme biosynthesis through modulation and, perhaps, multicellular eukaryotic organisms. of iron-sulfur cluster biogenesis. Knockdown of these novel Results: We assessed the effect of a calorie restriction diet and candidates in D.rerio embryos resulted in disruption of heme numerous mutations extending yeast life span on the age-related biosynthesis in all cases, confirming the functional predictions. alterations in proteomes and lipidomes of organelles involved The screening procedure also identified intriguing new in lipid metabolism. These organelles include the endoplasmic physiological roles for heme biosynthesis in immune response and reticulum (ER), peroxisomes and lipid bodies. We also examined hematological disorders. the spatiotemporal dynamics of organellar proteomes and Conclusions: This case study demonstrates that large-scale, lipidomes in long-lived yeast mutants impaired in various context-specific gene expression screening can reconstruct pathways of lipid metabolism. Implementing our understanding partially unknown biological pathways in mammals, while of the mechanism linking longevity and lipid metabolism, we simultaneously identifying the biological contexts where those developed a life-span assay that was used for a high-throughput pathways operate. screening of extensive compound libraries and revealed novel anti-aging small molecules. The effect of these molecules on DS1-1-27 organellar proteomes and lipidomes was elucidated. Conclusions: Our findings suggest a model for the mechanism Gender dependent progression of systemic metabolic linking longevity and lipid metabolism in the ER, peroxisomes states in early childhood and lipid bodies. In this model, a calorie-rich diet suppresses Ermolov, Andrey1; Sysi-Aho, Marko2; Nikkilä, Janne1; Seppänen- peroxisomal oxidation of free fatty acids (FFA) that originate from Laakso, Tuulikki2; Simell, Olli3; Kaski, Samuel1; Orešič, Matej2 neutral lipids synthesized in the ER and deposited within lipid 1Helsinki University of Technology, Department of Information bodies. The resulting accumulation of arrays of FFA within lipid and Computer Science, Espoo, Finland; 2VTT technical research bodies initiates several negative feedback loops regulating the centre of Finland, Systems Biology and Bioinformatics, Espoo, metabolism of neutral lipids. Due to the action of these negative Finland; 3University of Turku, Department of Pediatrics, Turku, feedback loops, aging yeast accumulate diacylglycerol (DAG). Finland The buildup of FFA and DAG triggers lipoapoptosis and activates a signal transduction network affecting multiple longevity-related Objectives: We introduce a novel concept suggesting that cellular targets, thereby shortening life span. We identified several children even at a very young age can be categorized in terms novel anti-aging small molecules that significantly delay yeast Posters of metabolic state as they advance in development. The Hidden aging by targeting lipid metabolism and remodeling organellar Dedicated Markov Models (HMM) were used as a method for discovering proteomes and lipidomes. the underlying progression in the metabolic state. We applied the methodology to study metabolic trajectories in children between DS1-1-29 birth and four years of age, based on a series of samples selected from a large birth cohort study. Metabolic pathway analysis in studying the evolution of Results: We found multiple previously unknown age- and gender- metabolism and the emergence of network properties related metabolome changes of potential medical significance. Ullrich, Alexander; Flamm, Christoph Specifically, we found that the major developmental state University Vienna, Vienna, Austria differences between girls and boys are attributed to sphingolipids. In addition, we demonstrated the feasibility of state based Objective: The emergence and evolution of metabolism, its alignment of personal metabolic trajectories. We show that parts -metabolites and enzymes-, as well as its properties on children have different development rates at the level of the network level, is an intriguing and still open question. We metabolome and thus the state based approach may be introduce an improved metabolic pathway analysis tool for the advantageous when applying metabolome profiling in search of application in a simulation of the functional evolution of ribozyme markers for subtle (patho)physiological changes. catalyzed metabolisms in a graph-based Toy-universe. During Conclusions: We found Hidden Markov Models as a natural the evolutionary process the tool is used to define a realistic choice to model the metabolic state progression, because HMMs selection criteria (e.g. metabolic yield) for the metabolisms, allow (1) intuitive evaluation of the most important metabolic leading to networks with properties as observed in their real factors characterizing different states as well as transitions world counterparts (e.g. small-world). Further it is used to analyze between the states, (2) alignment of multivariate metabolic less well understood properties of the resulting networks, such time trajectories for different individuals, and (3) modeling time- as robustness and modularity, to make predictions about their associated progression from a relatively small amount of data. origin and development. It can also be used as a stand-alone tool The presented computational framework may be also suitable for the enumeration of elementary modes from a stoichiometric for more complex study designs, e.g., when state changes are matrix and the subsequent computation of minimal cut-sets. searched for as indicators of disease development, or when Results: Our tool performs better in terms of memory efficiency interventions are launched to prevent or cure the disease. than known similar network analysis tools. It generates fewer candidates during the elementary modes enumeration due to a DS1-1-28 new row ordering. Fewer preliminary cut-sets have to be kept during the computation of the minimal cut-sets, since we use Using organelle proteomics and lipidomics for the a depth-first search instead of a breadth-first approach. The identification of novel anti-aging small molecules that connectivity distribution of the simulated networks approximates target lipid metabolism the power-law with increasing size and hub-metabolites originate Titorenko, Vladimir; Goldberg, Alexander; Gregg, Christopher; mostly from early stages. The robustness of the networks Boukh-Viner, Tatiana; Kyryakov, Pavlo; Bourque, Simon; Aziz, increases throughout evolution, even if fitness is not changing. Zeinab; Chang, Andrew; Cyr, David; Kayembe, Mulanda; Kim, Conclusions: From the above observations it can be suggested Hyun Young; Machkalyan, Gayane; Milijevic, Svetlana; Mudhar, that evolution of robustness is possible without even selecting for Ramandeep; Quashie, Peter; Ramlal, Nishant; Uscatescu, Victor; it and it can also be neutral. Further, it is fair to assume that the Askari, Mohammad Sharif way in which enzymes evolve in metabolic networks and thus the Concordia University, Biology Department, Montreal, Canada change in connectivity distribution accounts to some extent for the increase in robustness. Objective: The fundamental mechanisms of aging are conserved across phyla. Aging of multicellular eukaryotic organisms affects numerous processes within cells. We use the yeast

ICSB 2008 67 DS1-1-30 Results: Hierarchical clustering and clustering via self organizing maps were used in conjunction with gene ontology A kinetic model of the trehalose pathway in (GO) terminologies in order to determine which clusters were Saccharomyces cerevisiae significantly enriched for process and functional annotations. The Simeonidis, Evangelos; Messiha, Hanan; Spasic, Irena; transcriptome levels for qdr3Δ/qdr3Δ displayed differences from Smallbone, Kieran; Malys, Naglis; Kell, Douglas the other two mutants including the reference strain when the The University of Manchester, MCISB, Manchester, United carbon source was limited rather than the nitrogen source. Kingdom Conclusions: Of the down-regulated transcripts, several were observed to be associated with “vacuolar transport” and Objective: Trehalose is a disaccharide that has been extensively “sporulation” process GO terms. Up-regulated transcripts studied in baker’s yeast. Its apparent role is to function as a were observed to be significantly associated with amino acid carbohydrate reservoir, but it is also now thought to be a crucial biosynthesis, mitochondrial and ribosomal biogenesis, cofactor/ part of an important stabilizing mechanism for proteins and coenzyme metabolic processes and ergesterol biosynthesis. cellular membranes under stress conditions such as heat shock. The metabolic pathway that produces trehalose is believed DS1-1-32 to regulate glucose uptake, particularly when the cell exists in an adverse environment. For the above reasons, and also Growth-rate control of coding and non-coding RNAs in because of its numerous applications in the cosmetics, food and yeast pharmaceutical industries, trehalose has become an important Castrillo, Juan I.1; Hayes, Andrew2; Rash, Bharat2; Zeef, Leo A.2; biotechnological product. Hoyle, David3; Zhang, Nianshu4; Wardleworth, Leanne5; Oliver, Results: The trehalose pathway in yeast consists of a small Stephen G.4 number of reactions, but these are “arranged” in a metabolic 1University of Cambridge, Cambridge Systems Biology Centre, cycle and are governed by a highly complicated regulatory Biochemistry, Cambridge, United Kingdom; 2University of system of inhibitions and/or activations. Due to this complexity, Manchester, Faculty of Life Sciences. Michael Smith Building, the operation of the pathway is difficult to study experimentally. Manchester, United Kingdom; 3University of Manchester, We have built a kinetic model of the trehalose cycle, based on Northwest Inst. for Bio-Health Informatics (NIHBI), Manchester, Dedicated characterizations of the mechanism of each enzymatic reaction in United Kingdom; 4University of Cambridge, Cambridge Systems Posters the pathway. The model’s kinetic constants were collected with Biology Centre, Biochemistry, Cambridge, United Kingdom; the help of a text mining toolbox, KiPar, developed to retrieve 5University of Manchester, Faculty of Life Sciences. Michael Smith kinetic parameters of interest from publicly available scientific Building, Manchester, United Kingdom literature. Simulations are run for normal growth conditions and under stress, which is modelled by altering enzyme activities in Objective: Cell growth underlies many key cellular and accordance with observed gene expression changes. development processes, yet a limited number of studies have Conclusions: First, the results of the simulations are compared been conducted on cell growth regulation. Using chemostat to study the response variations among different growth culture, we measured the impact of changes in flux (growth rate) conditions. In addition, our model is evaluated against an existing on the transcriptome, proteome, endo- and exo-metabolomes canonical S-system model of the trehalose cycle in yeast [Voit of the yeast Saccharomyces cerevisiae (Castrillo et al., 2007. J 2003, J. Theor. Biol., 223, 55-78]. Our results demonstrate the Biol 6:4). The yeast YG_S98 array (Affymetrix, Inc.) contains 160 clear requirement for mathematical modelling approaches in the specific probe sets for the analysis of non-coding RNAs. We effort to decipher and understand the function and control of wanted to test the possibility that some small RNAs could also be biochemical pathways. subject to growth-rate regulation. Results: The results show clear growth-associated trends for DS1-1-31 a number of small nuclear RNAs (snRNAs), tRNAs, and small nucleolar RNAs (snoRNAs) involved in rRNA processing and Transcriptional response to the deletion of two multi-drug modification. Since non-coding RNAs do not exert their function resistance genes, QDR3 and PDR3 in Saccharamyces in isolation, we also studied the pattern of gene expression of cerevisiae non-coding RNAs and associated proteins. The results point to Kuzu, Guray1; Dikicioglu, Duygu1; Rash, Bharat2; Pir, Pinar3; the existence of coding and non-coding RNAs growth-regulated Hayes, Andy2; Oliver, Stephen3; Kirdar, Betul1 processes and a fine regulation of the expression of genes 1Bogazici University, Department of Chemical Engineering, specifying non-coding RNAs and their associated proteins during Istanbul, Turkey; 2University of Manchester, Faculty of Life cell growth. Sciences, Manchester, United Kingdom; 3University of Conclusions: The results point to the coordination of the Cambridge, Department of Biochemistry, Cambridge, United synthesis of non-coding RNAs and the proteins with which they Kingdom functionally interact in response to changes in growth rate. Our results have direct implications for a systems-level understanding Objective: In order to survive in the presence of chemical of growth control in the eukaryotic cell. compounds, cells have defense mechanisms so that they can respond to these unrelated compounds. Multi-drug resistance is DS1-1-33 one of these defense mechanisms in which membrane proteins play an important role. The membrane proteins that are key The gene-function relationship in the metabolism of S. molecules in this resistance phenomenon can be assigned Cerevisiae and digital organisms into two families, ABC (ATP-binding-cassette) and MFS (major Gerlee, Philip1; Lundh, Torbjörn2; Zhang, Bing3; Anderson, facilitator super-family) transport families. Qdr3p is a multidrug Alexander4 transporter of the major facilitator super-family, required for 1Niels Bohr Institute, Center for Models of Life, Copenhagen, resistance to quinidine, barban, cisplatin, and bleomycin; and Denmark; 2Chalmers and Gothenburg University, Department Pdr3p is a transcriptional activator of the pleiotropic drug of Mathematics, Göteborg, Sweden; 3Vanderbilt University, resistance network, regulates expression of ABC transporters Department of Biomedical Informatics, Nashville, United States; through binding to cis-acting sites known as PDREs (PDR 4University of Dundee, Division of Mathematics, Dundee, United responsive elements). In order to gain further insights into Kingdom the elucidation of the molecular mechanisms leading to drug resistance in yeast, the genome wide expression levels as well as Objective: Many biological systems form structures which can exo- and endo-metabolome levels of qdr3Δ/qdr3Δ and pdr3Δ/ best be viewed as networks consisting of a set of nodes and links pdr3Δ of the yeast Saccharomyces cerevisiae together with hoΔ/ connecting the nodes. Many of these networks exhibit a scale- hoΔ under different nutrient limitations were analyzed at steady free degree distribution and therefore deviate from the classical state. description of complex networks which predicts an exponential

68 ICSB 2008 degree distribution. We have studied the metabolic gene-function unknown ΔG0 value of a reaction from known ΔG0 values of network in yeast and digital organisms from the artificial life chemically similar reactions. platform Avida. The gene-function network is a bipartite network Results: To quantify the chemical similarity of biochemical in which a link exists between a gene and a function (pathway) if reactions we have established a detailed classification procedure that function depends on that gene, and can also be viewed as a that assigns 3304 different chemical attributes to atomic groups decomposition of the more traditional functional gene networks, occurring in presently characterized biochemical metabolites. where two genes are linked if they share any function. Changes in these attributes between the substrate and product Results: We show that the gene-function network exhibits two molecules are tracked on a per-atom basis and similarities distinct degree distributions: the gene degree distribution is scale- between these reaction-specific attribute changes are assessed free while the pathway distribution is exponential. This is true for by the Tanimoto coefficient (T) assuming values between 0 both yeast and digital organisms which suggests that this is a (complete dissimilarity of reactions compared) and 1 (identity general property of evolving systems. of reactions compared). Testing our method across a set of Conclusions: One possible explanation for this structure is 1546 biochemical reactions 216 of which being covered by that in the network the genes acquire new links according to experimentally determined ΔG0 values - the root-mean-square preferential attachment while the pathways receive new links distance (RMSD) between predicted and measured ΔG0 values independent of their degree. Gene duplication, which is the main amounted to 8.0 kJ/mol, if a minimum similarity of T>0.6 to mechanism by which new genes are created, acts as preferential reactions with known ΔG0 values is assumed. attachment for the pathways, but the exponential pathway degree Conclusions: This value is significantly smaller than the distribution suggests that the rate of functional divergence is RMSD of 10.5 kJ/mol achieved with the commonly used group considerable. The duplication of pathways/functions could on contribution method. However, for less similar reactions, the the other hand explain the scale-free distribution of the genes, group contribution method produces a more accurate predictions and measuring the overlap between different pathways, in terms and a combination of both approaches is proposed. Clustering of the genes which constitute them, showed that this is a likely all reactions of a given metabolic network according to chemical mechanism in yeast evolution. In conclusion we have presented similarity allows to identify minimal sets of reactions for which ΔG0 a new way of analysing the gene-function dependence which values yet have to be experimentally determined in order to make sheds new light on the evolution of genes and functionality, and reliable predictions of ΔG0 values for the remaining reactions. especially the importance of function duplication in evolution. DS1-1-36 DS1-1-34 Mapping regulatory mechanisms of Saccharomyces

Kinetic model of the Plasmodium falciparum glycolytic cerevisiae lipid metabolism Posters pathway during the asexual phase Nookaew, Intawat1; Laoteng, Kobkul2; Thammarongtham, Dedicated Penkler, Gerald; Snoep, Jacky Chinae2; Meechai, Asawin3; Cheevadhanarak, Supapon4; University of Stellenbosch, Stellenbosch, South Africa Bhumiratana, Sakarindr5; Nielsen, Jens6 1King Mongkut’s University of Technology, Thonburi, Pilpt Plant Objective: Malaria infection results in the death of over 1 million Development and Training Institute, Bangkok, Thailand; 2National children annually. Furthermore, even more lives may be lost Center for Genetic Engineering and Biotechnology (BIOTEC), unless novel drugs are discovered to combat the emergence Bangkok, Thailand; 3King Mongkut’s University of Technology of resistance to current front line drugs. The malarial parasite, Thonburi, Department of Chemical Engineering, Bangkok, Plasmodium, is entirely dependent on the glycolytic pathway Thailand; 4King Mongkut’s University of Technology Thonburi, for energy and the path is thus a prime drug target. Although School of Bioresources and Technology, Bangkok, Thailand; many of the individual glycolytic enzymes have been studied, the 5National Science and Technology Development Agency, Ministry pathway as a whole has not been considered. We thus aimed of Science and Technology, Bangkok, Thailand; 6Chalmers 1) to construct a kinetic model for the Plasmodium falciparum University of Technology, Department of Chemical and Biological glycolytic pathway during the parasite asexual phase and 2) to Engineering, Gothenburg, Sweden use metabolic control analysis to quantify the control of each enzyme on the glycolytic flux and intermediate steady state Objective: There is a rapidly increasing demand for poly- concentrations. unsaturated lipids but the supply has not been sufficient to meet Results: Using the experimentally determined kinetic properties the market demand. A promising source of lipids is through for each enzyme and the theoretical frameworks available for microbial production, especially from fungi such as Mortierella analysing kinetic pathways we are able to show which enzymes sp. and Mucor sp. Yet the regulation and control mechanisms of are most important in controlling glycolytic flux and steady state the lipids production in these organisms are not well understood. intermediate metabolite concentrations. In this study Saccharomyces cerevisiae was used as model Conclusions: This work, which identified key glycolytic enzymes, organism to study the regulation of lipid biosynthesis in these has pin pointed potential drug targets within the Plasmodium fungi. falciparum glycolytic pathway and may thus be important for drug Results: Transcriptome data from various growth conditions that design and the search for novel antimalarial compounds. affected lipid composition were mapped on interaction network related with lipid metabolism (977 nodes and 7395 edges). Using DS1-1-35 a heuristic search algorithm of transcriptome mapped network, highly significant subnetworks were identified. Many regulation Gibbs free energy changes of biochemical reactions modules in the subnetworks were elucidated and compared with inferred from reaction similarities literature data on sterol uptake, beta-oxidation, desaturation and Rother, Kristian1; Hofmann, Sabrina2; Bulik, Sascha2; Hoppe, sphingolipid related module. Andreas2; Holzhuetter, Herrmann-Georg2 Conclusions: More insight understanding into the control and 1International Institute of Molecular and Cell Biology, Warsaw, regulatory mechanism of lipid metabolism in S. cerevisiae was Bioinformatics and Protein Engineering Lab, Warsaw, Poland; revealed in this study. The high confidence regulation networks 2Institute of Biochemistry, Charite Universitätsmedizin Berlin, were validated by known regulatory mechanisms from the Computational Biophysics Group, Berlin, Germany literature. The highly significant subnetworks can be used to guide genetic modifications for further strain improvement of fungi to Objective: An indispensable prerequisite for the thermodynamic increase the production of complex lipids. and kinetic modeling of biochemical reaction networks is to assign a reliable value for the standard Gibbs free energy change

(ΔG0) to each reaction and transporter. However, for genome- wide metabolic networks experimental ΔG0 values are scarce. Here we propose a novel computational method to infer the

ICSB 2008 69 DS1-1-37 quantified as the intracellular carbon fluxes using13 C-labeling experiments. Key findings include that the largest relative A single molecule approach to stringent response in concentration changes were observed for the fructose-1,6- individual living bacterial cells bisphosphate pool. English, Brian; Elf, Johan; Leroy, Prune Conclusions: Network-embedded thermodynamic integration Uppsala University, Cell and Molecular Biology, Uppsala, Sweden of metabolic fluxes and metabolite concentrations indicated that Objectives: The stringent factor RelA binds to a small fraction sites of metabolic flux regulation differ with condition. Fructose- of ribosomes, where it synthesizes the global transcriptional bisphosphate aldolase and pyruvate kinase were predicted to be regulator ppGpp in response to amino acids deprivation. Our the most likely targets in glycolysis for regulation under glycolytic objective is to study the binding kinetics of individual RelA conditions, whereas triose-phosphate isomerase appears to be molecules to the ribosome in living cells and to observe how its the key control point during gluconeogenesis. Throughout central kinetics changes during a nutritional down-shift. carbon metabolism, it was examined whether these predictions Results: We present a single-molecule fluorescence assay to were verified by the proteomics data. directly observe the stringent response in individual living E. coli cells. We have assembled a state of the art confocal excitation/ DS1-1-39 wide-field detection fluorescent microscope for tracking of the fluorescently labeled RelA proteins. Colinearity of gene order and enzymatic steps in metabolic For this purpose, we have created fusions of RelA with a operons of Escherichia coli photo-activatable fluorescent protein. WhileE. coli contains Kovacs, Karoly1; Hurst, Laurence2; Papp, Balazs1 on average about 100 RelA molecules and 10000 ribosomes, 1Biological Research Center, Evolutionary Systems Biology using a photo-activatable fluorescent probe we can activate Group, Szeged, Hungary; 2University of Bath, Department of only a few fluorescent RelA molecules per cell at any given Biochemistry, Bath, United Kingdom time. Furthermore, short confocal excitation pulses allow us to selectively activate fluorophors in a small region of a living Objective: It is well established that gene order in prokaryotic bacterial cell. We induce stringent response by rapid addition genomes is not random. Genes within the same operon often of amino acid hydroxamates. Since our fluorescent tag is code for enzymes involved in the same metabolic pathway or Dedicated photoconvertible, we can repeat the RelA tracking experiments proteins that belong to the same complex. However, it is almost Posters many times in the same E. coli cell, both before and after completely unexplored if intra-operonic gene order is random. triggering stringent response. The continuous replenishing of Here we ask if gene order within metabolic operons reflects the fluorescence by photo-conversion considerable extends the functional order of the encoded enzymes acting in the same experimental timescale that previously was limited to a few frames biochemical pathway (colinearity). We hypothesize that colinear by photobleaching in conventional single molecule tracking arrangement might be adaptive by enabling the temporal ordering assays. of enzymes in linear metabolic pathways (i.e. upstream operonic Conclusions: We observe binding of individual RelA to the genes are expressed earlier than downstream ones). ribosome in living cells. We present trajectories of individual RelA Results: Using a simple stochastic kinetic model of gene molecules diffusing in living E. coli cells with a time resolution expression and metabolic pathway operation we not only show of 10 milliseconds and a spatial precision of 50 nanometers. that a colinear arrangement can increase pathway productivity, The high resolution of the experiments makes it possible but also find that the advantage of colinearity is greater in lowly to characterize RelA binding kinetics under varying growth expressed operons. To see if these theoretical predictions are conditions. upheld, we investigated the metabolic operons and biochemical pathways of Escherichia coli. Our analyses showed that, on DS1-1-38 average, colinearity of intra-operonic gene order is significantly higher than expected by chance. Furthermore, in accordance with Identification of key metabolic regulation sites in our predictions, we found that colinearity is more pronounced for Saccharomyces cerevisiae through thermodynamic lowly expressed operons and for genes that are not adjacent in modeling of -omics data the operon. Costenoble, Roeland; Picotti, Paola; Aebersold, Ruedi; Sauer, Conclusion: Our study on E. coli metabolic operons provides Uwe the first systematic evidence that intra-operonic gene order is not ETH Zuerich, Institute of Molecular Systems Biology, Zuerich, random, and we conclude that temporal ordering of enzymes in Switzerland metabolic pathways imposes some constraints on gene order evolution even within operons. Objective: Although the topology of the metabolic network of Saccharomyces cerevisiae is known, its regulation is not well DS1-1-40 understood. Elucidation of systemic properties of metabolic regulation, for instance in response to changing environmental Understanding translation regulation in yeast using conditions, requires experimental data on the network correlations between mRNA levels and protein levels components but also their interpretation within quantitative Olivares, Roberto1; Usaite, Renata2; Nielsen, Jens1 (mathematical) models. Most mathematical models of metabolic 1Chalmer Universtity of Technology, Department of Chemical and regulation are, however, limited by the completeness, the Biological Engineering, Göteborg, Sweden; 2Center for Microbial quantitative accuracy and the restricted diversity of the data Biotechnology, DTU-Biosys, Lyngby, Denmark they are based upon. Here, we generated comprehensive and quantitative data sets of metabolic proteins, metabolites and Objective: The aim of this study is to perform a comparative metabolic fluxes in the central carbon metabolism ofS. cerevisiae analysis using high quality proteome and transcriptome data under five different environmental conditions, and used different taken from literature. Mapping the correlation between mRNA mathematical methods to analyse these data. and protein levels we clustered genes considering the distance Results: Based on a genome-scale model of yeast from the perfect correlation and understand how the cell metabolism, targets were selected that were to be detected regulate the translation under different environmental and genetic by a novel, targeted proteomics method based on multiple- perturbations. reaction monitoring (MRM). By this method, we were able Results: Posttranscriptional regulation causes the deviation of a to near comprehensively quantify all metabolic enzymes in perfect correlation between mRNA level and protein level. In an central metabolism (ca. 150), including most isoenzymes, at effort to unravel global pattern in cell response to this control, we an unprecedented sensitivity. Additionally, we quantified by use hypergeometric test to find biological process to those genes LC-MS/MS and GC-TOF analysis absolute concentrations that fall close to the correlation one to one. Mapping mRNA level of 60 intracellular metabolites, including redox cofactors and and Protein level we clustered genes using a simple Euclidian nucleotides. Finally, the functional output of the network was distance considering 10% of deviation from a perfect relation one-

70 ICSB 2008 to-one. This clustering draws a region where there is no regulation genome was performed and based on this a genome- at translational level. We found that the GO process categories scale metabolic model of A. oryzae was reconstructed. The represented through all datasets were carboxylic acid metabolic metabolic network and formulated model was validated with process, organic acid metabolic process, amino acid biosynthetic experimental and literature data from fermentation process as process and nitrogen compound biosynthetic process. To have well as transcription analysis, accurately predicting biomass more information we found the pathways that are linked to those yield and enzyme yield on different carbon sources. It serves as genes enriched in different categories. We can see that amino an important resource for gaining further insight into A. oryzae acid synthesis pathway, glycolytic pathway and TCA cycle and physiology. are mainly regulated at translational level when the cell is under different stress conditions. DS1-1-43 Conclusions: Using transcriptome and proteome data for yeast Saccharomyces cerevisiae we found that the mechanism Saccharomyces cerevisiae functional genomics applied to for translation regulation in genes involved in central carbon nutraceutical development metabolism and amino acid biosynthetic pathways is conserved Papadakis, Manos A.1; Otero, José Manuel2; Nielsen, Jens2; under different carbon sources and genetic perturbations. In order Panagiotou, Gianni1 to find if this global response is evolutionary conserved in the cell, 1Technical University of Denmark (DTU), Department of as future work, we will extent the same analysis including data Systems Biology, Kgs. Lyngby, Denmark; 2Chalmers University from other organisms. of Technology, Department of Quantitative Systems Biology, Gothenburg, Sweden DS1-1-41 Objective: Industrial biotechnology offers the opportunity to mRNA stability coordinates the unfolding of gene develop novel processes using microbial cells for the production expression in the long-period yeast metabolic cycle of high added-value compounds e.g., nutraceutical compounds. Altafini, Claudio1; Soranzo, Nicola1; Zampieri, Mattia1; Farina, Ferulic acid is an abundant cinnamic acid derivative found Lorenzo2 in the plant cell wall. Food supplementation with ferulic acid 1SISSA Int. Sch. for Advanced Studies, Trieste, Italy; 2Univ. of is considered a novel nutritional approach to reduce cellular Rome, Rome, Italy oxidative damage and other pathological events occurring in diverse neurological disorders. However, the exact mode of Objective: In yeast, genome-wide periodic patterns associated action of antioxidant compounds in human cells has yet to be with energy-metabolic oscillations have been shown recently for mechanistically elucidated. both short (approx. 40 min) and long (approx. 300 min) periods. Results: CEN.PK 113-5D yeast strain has been physiologically Posters The main point of this work is to shown that in the “long characterized in chemostat cultivations under aerobic and Dedicated period yeast metabolic cycle” there is a direct proportionality anaerobic conditions while gene expression and metabolome between the phase at which the periodic genes peak and the data have been generated when the strain was cultivated in the corresponding mRNA turnover rates. presence of ferulic acid. After extensive microarray data analysis, Results: It is shown that for periodic genes, arranged in classes 106 genes were found to be significantly differentially expressed according either to expression profile or to function, the pulses with ferulic acid supplementation, independent of aerobic or of mRNA abundance have phase and width which are directly anaerobic conditions. Furthermore, 60 of the 106 gene products proportional to the corresponding turnover rates. In terms of were utilized for the construction of a ferulic acid specific protein- dynamical models, such a progressive broadening and smoothing protein interaction network based on information in SGD and of the response to a sequence of (transcriptional) pulses can STRING databases. be described by means of simple linear input-output models of The integration of the available expression data into the network increasing order having ``low-pass’’ characteristics. and analysis of its topology suggested 16 gene products as key Conclusions: This ordered pattern reproduces faithfully the nodes modulated by ferulic acid. After a YEASTRACT-mediated cascade of events that constitutes the gene expression program survey for the transcription factors regulating the expression of of the organism in response to a burst of transcriptional activation. the corresponding 16 genes, and the subsequent transcription regulatory network inference, a biological signature reflecting the DS1-1-42 oxidative status of the cell was proposed. Conclusions: Systems biology tools in orchestration with Constructing Aspergillus oryzae as an efficient cell factory quantitative physiological characterization and molecular biology for protein production techniques have been applied to identify regulatory nodes Vongsangnak, Wanwipa1; Olsen, Peter2; Hansen, Kim2; affected by a model antioxidant compound. In order to elucidate Krogsgaard, Steen2; Nielsen, Jens1 holistically the mechanistic details of ferulic acid in yeast and 1Current address: Chalmers University of Technology, Department propose a similar model in humans a number of mutants have of Chemical and Biological Engineering, Gothenburg, Sweden; been constructed. The characterization of these mutants using a 2Novozymes A/S, Bagsvaerd, Denmark systems level approach under carbon-limited and nitrogen-limited conditions in batch cultures has been performed. Objective: Historically, the filamentous fungusAspergillus oryzae has been used in industrial fermentation production of various DS1-1-44 food supplements, and more recently as a cell factory for protein production. The genome database contains 12,074 annotated Modeling of the cross-talk between cholesterol genes, but the number of hypothetical proteins accounts for more homeostasis and drug metabolism in human primary than 50% of the annotated genes. Consequently, it is valuable to hepatocytes improve the annotation and integrate genomic, biochemical, and Rozman, Damjana1; Juvan, Peter1; Rezen, Tadeja1; Monostory, physiological information. Katalin2; Pascussi, Jean-Marc3; Belic, Ales4 Results: Here we improved the gene prediction and function 1University of Ljubljana, Faculty of Medicine, Centre for Functional assignment by our constructed large set of assembled Expressed Genomics and Bio-Chips, Ljubljana, Slovenia; 2Hungarian Sequence Tags (ESTs) and different bioinformatics techniques, Academy od Sciences, Budapest, Hungary; 3INSERM, respectively. We identified 1,046 newly predicted genes and Montpellier, France; 4University of Ljubljana, Faculty of Electrical assigned putative functions to 1,469 hypothetical proteins of the Engineering, Ljubljana, Slovenia existing genome database. Using the improved annotation we reconstructed the metabolic network of A. oryzae, resulting in 729 Objective: To understand the cross-talk between cholesterol enzymes, 1,314 enzyme-encoding genes, 1,073 metabolites and homeostasis and drug metabolism in the liver we employed 1,846 (1,053 unique) biochemical reactions. experimental approaches together with differential equation Conclusions: A much enhanced annotation of the A. oryzae mathematical modeling/simulation and the Bayesian inference of

ICSB 2008 71 gene interactions. DS1-1-46 Results: Transcriptional changes were measured by a dedicated Steroltalk microarray at 12, 24 and 48h on rifampicin, statins and Biologically realized bacterial flux distributions are original hypolipidemics treated human primary hepatocytes from optimal within a multi-objective space under different 7 individuals. Sterols and selected proteins have been measured environmental conditions at 48h by GC-MS or western blots. A mathematical model was Schuetz, Robert1; Kuepfer, Lars2; Zamboni, Nicola1; Heinemann, used to simulate gene expression measurements and Bayesian Matthias1; Sauer, Uwe1 inference to construct gene interaction networks from the 1Institute of Molecular Systems Biology, Zurich, Switzerland; simulated data. The simulated networks were compared to the 2Bayer Technology Services, Leverkusen, Germany network from the measurements. Various structural changes of the mathematical model have been considered and the model has Objective: Single objective functions within a flux-balance been adjusted to improve correspondence to the experimental framework cannot describe metabolic flux distributions under all data. The Bayesian networks obtained from the measured and environmental conditions. However, empirically chosen objectives the simulated data both show that expression of cholesterogenic have been found to make biologically meaningful predictions. A genes can not be predicted only from the expression of the major key question thus is whether one can understand the evolutionary regulatory factor SREBF2 but the expression of 3 additional genes optimization in terms of a combination or trade-off between is required. Networks also indicate a strong gene expression individual objectives. Here we address this question and attempt interaction between SREBF2 and CYP51A1, but not between to quantify such trade-offs. SREBF2 and HMGCR, suggesting regulation of HMGCR by other Results: Using a detailed stoichiometric model of Escherichia factor(s). coli central metabolism, we computed the flux cone that includes Conclusion: We have demonstrated how computational tools all valid network states with elementary-mode analysis. With a can be combined to gain novel perspectives of a biological multi-objective framework we then quantified thePareto surface, system of interest. A large number of perturbations of the system the subspace with all Pareto optimal flux solutions, for which the is critical for identification of gene interactions and differences value of one objective can only be increased by decreasing the in gene expression levels between individuals pose a serious value of others. In particular, we investigated the objective space problem. The proposed combination of modeling approaches spanned by the three competing objectives of a maximal ATP and Dedicated is general, but could be further improved by inclusion of the biomass yield and a minimal intracellular flux, which we identified Posters proteome and metabolome data. A larger number of individuals to be most relevant for growth on glucose. with measured expression profiling and SNP genotyping would be Next, we estimated the locations in the objective space of 48 in required to allow for modeling of inter-individual differences in the vivo flux distributions fromEscherichia coli grown under aerobic, measured data. anaerobic and nitrate-respiring batch conditions and chemostat Acknowledgement: This work is performed in the context of the conditions with varying dilution rates. Interestingly, while all STEROLTALK project contract No. LSHG-CT-2005-512096 under biologically realized flux states locate very close to different areas the EU 6th Framework Programme for RTD. of the Pareto surface, randomly generated fluxes do not behave in such an optimal way, but locate far away from the Pareto surface. DS1-1-45 Conclusions: Our results demonstrate that in vivo flux states are the result of an evolutionary optimization process rather Co-regulation of metabolic genes is better explained by than a chance event. Furthermore, depending on the prevailing flux coupling than by network distance environmental condition, different Pareto optimal combinations Notebaart, Richard1; Teusink, Bas1; Roland, Siezen1; Papp, are realized in vivo. As a case in point, while slowly growing Balázs2 chemostat conditions favor a high ATP yield, during aerobic batch 1Radboud University Nijmegen, NCMLS, Nijmegen, Netherlands; conditions, on the other hand, a high biomass yield and a low 2Institute of Biochemistry, Biological Research Center, Szeged, intracellular flux are dominating. Hungary DS1-1-47 Objective: To what extent can modes of gene regulation be explained by systems-level properties of metabolic networks? Systems analysis of metabolism in a minimal bacterium Prior studies on co-regulation of metabolic genes have mainly Yus, Eva1; Maier, Tobias2; Mourao, Andre3; Raineri, Emanuele2; focused on graph-theoretical features of metabolic networks Guell, Marc2; Sattler, Michael3; Serrano, Luis2; Wodke, Judith2 and demonstrated a decreasing level of co-expression with 1Centre for Genomic Regulation, Systems Biology, Barcelona, increasing network distance, a naïve, but widely used topological Spain; 2Centre for Genomic Regulation, Barcelona, Spain; index. Others have suggested that static graph representations 3Institute of Structural Biology, Munich, Germany can poorly capture dynamic functional associations, e.g. in the form of dependence of metabolic fluxes across genes in the Objectives: to understand the metabolism of Mycoplasma network. Here, we systematically tested the relative importance pneumoniae at a quantitative level. For this purpose, metabolism of metabolic flux coupling and network position on gene co- will be reconstructed based on homology searches and literature regulation, using a genome-scale metabolic model of Escherichia mining, based on which a synthetic medium will be designed. coli and Saccharomyces cerevisiae. This medium will be use as a proof of principle of the theoretical Results: After validating the computational method with empirical metabolic pathways. Substrates will be added labeled with data on flux correlations, we confirm that genes coupled by their carbon-13 and metabolic fluxes will be determined. enzymatic fluxes not only show similar expression patterns, but Results: a metabolic map of M. pneumoniae has been built. also share transcriptional regulators and frequently reside in the Gaps in different pathways have been filled manually by extracting same operon. In contrast, we demonstrate that network distance information from operon structure, phylogenetic analysis and data per se has relatively minor influence on gene co-regulation. on enzymatic activities determined experimentally. Moreover new Moreover, the type of flux coupling can explain refined properties pathways such as anaplerotic reactions or the use alternative of the regulatory network that are ignored by simple graph- carbon sources has been proposed. Based on this map a theoretical indices. synthetic medium for this bacterium has been formulated, and the Conclusions: Our results illustrate that modes of gene co- ability to support growth has been established. Furthermore this regulation can be better explained by a biochemically well- medium has been used to validate the map, and determining the grounded flux correlation based measure, than by network minimal components, for instance, the ability of cytidine to supply distance. Our work has important implications for comparative all other pyrimidines. NMR studies with 13C-glucose have been genomics and gene function predictions. In particular, the used to determine the fate of glucose carbons. We observe two concept of flux coupling could form the basis to test the reliability mayor fates of glucose, to lactate and acetate as end products, of predicted functional interactions by genomic context or high- with minor contribution to nucleosugar formation. Furthermore throughput functional genomics data. expression profiling by microarrays is used to analyze the levels of

72 ICSB 2008 metabolic enzymes under different carbon sources or metabolic The putative enzymes are purified and assayed for the specific stresses. reaction(s) using mass spectrometry based assays. Using Conclusions: metabolism of one of the smallest genome bearing this approach, several candidate reactions are being targeted organism has been reconstructed and tested experimentally. This including a putative deacetylase activity as well as candidate has fueled the development of a defined medium that is a tool to activities involved in butanoate metabolism. perform metabolic flux analysis. Conclusions: The generic approach is broadly applicable and allows to link uncharacterized ORFs to “missing” metabolic DS1-1-48 activities.

A multi-scale model of human skeletal muscle: extracting DS1-1-50 metabolite dynamics at cellular scale from measurements at whole organ scale Metabolite networks as a dynamic response analysis tool Schmitz, Joep; Jeneson, Jeroen; Nicolay, Klaas; Hilbers, Peter; for time series stress experiments in Escherichia coli van Riel, Natal Jozefczuk, Szymon; Catchpole, Gareth; Willmitzer, Lothar Eindhoven University of Technology, Biomedical Engineering, Max Planck Institute of Molecular Plant Physiology, Potsdam- Eindhoven, Netherlands Golm, Germany

Objective: Insight in the intracellular metabolic network Objective: Systems biology may be called the missing link underlying ATP synthesis and hydrolysis is essential to understand between molecular biology and physiology by integrating different skeletal muscle physiology and pathophysiology, such as levels of information about a biological system using efficient insulin resistance. However, skeletal muscle tissue is highly bioinformatics tools. Today one of the greatest challenge in the heterogeneous and dynamic. It contains three distinct muscle post-genomic era is extracting useful information about cellular fiber types with different metabolic characteristics. An additional processes based on dynamic networks models. Constant level of organization is the grouping of fibers of the same type development of non-targeted high-throughput metabolite profiling in motor units, enervated by the same motor neuron. Motor unit methods allows the measuring, identifying, and quantifying of up recruitment is a sophisticated, dynamic process. We developed to 200 metabolites in Escherichia coli using gas chromatography - a multi-scale model of human skeletal muscle. Model simulations mass spectrometry (GC-MS). Metabolites, as both substrates and were validated with in vivo 31P NMR measurements. products of metabolic reactions, as well as active regulators of Results: At the fiber level the model is composed of detailed cellular processes, can be regard as an interesting starting point models of mitochondria, glycolysis, ATP buffering and ATP to study biological systems. consumption. A set of fiber type specific parameters was Recent development of efficient bioinformatics tools enables us Posters identified and parameterized for the three different types of muscle to construct and study metabolite networks. Assuming that the Dedicated fibers. The whole muscle scale was modeled as the average observed pattern of correlations between metabolites reflects behavior of a representative pool of motor units of different the underlying biochemical network, metabolite - metabolite types. Dynamic data of [PCr], [Pi] and [ATP] were obtained from correlations can provide valuable information about unknown 31P NMR spectra measured during in magnet bicycle exercise. pathways and mechanisms of cellular regulation. The identification The mitochondrial enzyme content in each of the three types of of network topology can give us ideas about the organization muscle fiber was estimated based upon the31 P NMR data. The of whole cellular systems. Finally, network analysis can illustrate predicted rate of post-exercise [PCr] recovery was the highest in the dynamics of biological systems in response to specific oxidative slow twitch fibers, intermediate in oxidative fast twitch perturbations. Here we present a systems biology study of the fibers and the lowest in glycolytic fast twitch fibers. These [PCr] kinetic response of Escherichia coli to three different stress recovery rates agreed well with values available in literature. conditions and metabolic shift. Conclusion: The computational model is able to reproduce We used GC-MS profiling to study the metabolic response of human skeletal muscle dynamics as obtained with 31P NMR Escherichia coli to heat shock, cold shock, oxidative stress, and spectroscopy and predicts the behavior of different types of glucose to lactose utilization. For all experiments, samples for skeletal muscle fibers.31 P NMR spectroscopy is a powerful metabolites and RNA measurement were taken minimally at 11 research and clinical tool to study in vivo dynamics of skeletal timepoints. Here we describe the metabolic networks, describe muscle metabolism. However, the 31P NMR data represents the network topology, and metabolite-metabolite correlations behavior of the whole muscle and, thus, the mean behavior of characteristic for these different stress conditions. all motor units / muscle fibers, hampering any straightforward Results: will be provided interpretation of the data. The developed multi-scale model can Conclusion: will be provided be applied to extract the dynamic behavior of different types of myocytes from 31P NMR measurements. DS1-1-51

DS1-1-49 Singing in concert. Couplig of glycolytic oscillations with oscillations in mitochondrial membrane potential in Filling metabolic pathway gaps using generic mass Saccharomyces cerevisiae spectrometry assays Olsen, Lars Folke; Poulsen, Allan Korsgaard; Brasen, Jens Robert, Martin; Saito, Natsumi; Soga, Tomoyoshi; Tomita, Masaru Christian; Andersen, Ann Zahle Keio University, Institute for Advanced Biosciences, Tsuruoka, University of Southern Denmark, Department of Biochemistry and Japan Molecular Biology, Odense, Denmark

Objective: The complexity of metabolic systems even in the Objective: In a dense suspension of non-growing yeast cells simplest organisms is a formidable challenge. An accurate model oscillations in glycolysis and mitochondrial membrane potential of metabolism can allow to better understand cell function but can be observed under anaerobic or semi-anaerobic conditions. requires a complete description of metabolic activities. Emerging The occurrence of oscillations is a collective phenomenon: metabolomics datasets suggest the presence of numerous and A single cell shows no or only damped oscillations, while a unsuspected gaps in existing reconstructions of the metabolic suspension containing approximately 109 cells may show networks of model organisms such as E. coli. In an attempt to fill- sustained oscillations. We have investigated the coupling between in those gaps we are using computational predictions and simple oscillations in glycolysis and mitochondrial membrane potential in vitro assays using mass spectrometry to monitor enzyme by simultaneous measurements of NADH fluorescence and activity. fluorescence of the membrane potential-sensitive carbocyanine Results: Unassigned activities in a metabolic network are dye DiOC2(3), that almost exclusively measures the mitochondrial identified and matching enzyme candidates selected based on membrane potential (MMP). sequence characteristics and/or available functional information. Results: We found that NADH and MMP oscillate in phase. We

ICSB 2008 73 further observed that inhibitors of glycolytic enzymes, such as DS1-1-53 hexokinase (2-deoxyglucose) or aldolase (iodoacetate), inhibit oscillations in both NADH and MMP. As opposed to this, low Constraint based motif analysis concentrations of uncouplers of oxidative phosphorylation (e.g. Kremling, Andreas FCCP) dissipate the membrane potential, but have little or no Max-Planck-Institute, Magdeburg, Germany effect on the oscillations in NADH. Higher concentrations of uncouplers also inhibit the NADH oscillations. Inhibitors of ATP Objectives: The understanding of the interplay of many hydrolysis mediated by F0F1-ATPase (e.g. azide) strongly reduce components in complex biochemical reaction networks requires the amplitude of both NADH and MMP oscillations. Furthermore, analytical tools based on a sound mathematical description of inhibitors of the plasma membrane H+-ATPase (PMA1) (e.g. system. In recent years, carbohydrate uptake and metabolism in Omeprazole) also reduce the amplitudes in oscillations. Finally, E.coli has attracted attention from experimental and theoretical addition of atractyloside, an inhibitor of the mitochondrial ADP/ working groups (Mettetal, PNAS, 2006, Nishio, Mol. Sys. Bio., ATP antiporter, also reduces the amplitudes of the oscillations of 2008) since the biological knowledge is rich and can be used for both NADH and MMP. model set up and analysis. Conclusions: We conclude that oscillations in glycolysis control In previous studies, different mathematical models were set up the oscillations of membrane potential through the F0F1-ATPase and analyzed to describe the relationship between the specific and the mitochondrial ADP/ATP antiporter. Furthermore, the growth rate and a cellular sensory system which measures the ability of the cells to show glycolytic oscillations is determined by glycolytic flux and which is connected to the global transcription the ATP/ADP ratio in such a way that at very high or low ATP/ factor Crp (Kremling, BMC Syst. Bio., 2007, Kremling, Bioinf., ADP ratios no oscillations will occur. A high ATP/ADP ratio can 2008). In the contribution at hand, the approach is generalized be obtained through inhibition of the F0F1 ATPase or the plasma with respect to uncertain and incomplete knowledge on kinetic membrane H+-ATPase (low ATP hydrolysis rate), while a low ATP/ rate laws. ADP ratio can be obtained by addition of high concentrations of Result: A novel approach “constrained based motif analysis” e.g. FCCP (high ATP hydrolysis rate). is introduced that comprises the following steps: (i) Definition of a metabolic network by a stoichiometric matrix; (ii) Definition DS1-1-52 of a cellular measurement (sensory) system and (iii) definition of Dedicated physiological input parameters like the specific growth rate. The Posters An integrative and comparative study on regulation of approach requires no assumption on the choice of the kinetic rate yeast metabolism by two major nutrients; glucose and law, but the signs of the derivatives with respect to the involved ammonium components are fixed. The approach calculates the concentration Dikicioglu, Duygu1; Rash, Bharat2; Pir, Pinar3; Hayes, Andy4; Oliver, of the metabolites dependent from the input and is suitable for Stephen G.3; Kirdar, Betul1 networks of moderate complexity that show a certain structure 1Bogazici University, Chemical Engineering, Istanbul, Turkey; that can be found more frequent in biochemical networks than 2University of Manchester, Faculty of Life Sciences, Manchester, others. Such subnetworks are also called motifs. The approach United Kingdom; 3University of Cambridge, Department of was applied for the central metabolism of E. coli and it is shown Biochemistry, Cambridge, United Kingdom; 4University of that a feed-forward loop in the network is identified as such a Manchester, Faculty of Life Sciences, Manchester, United motif that shows an advantage over other possible network Kingdom structures: the feed-forward loops guarantees robustness while a different structure tends to oscillating behavior. Objective: External factors cause a dynamic response in the Conclusions: The approach allows to explore and to access gene products of an organism. This response is carried out network structures with respect to network characteristics, e.g. through a cascade of interactions involving transcription factors robustness, without the detailed knowledge of the kinetic rate as well as a large number of proteins. The Dynamic Regulatory laws. Events Miner (DREM) is a software tool based on input-output hidden Markov models to model regulation in organisms using DS1-1-54 time-series gene expression data together with ChIP-chip or motif data. The tool basically aims to identify bifurcation points Studies on the mating system of Biomphalaria alexandrina corresponding to places in time series where the expression of a snails subset of genes diverges from the rest of the genes. In this study, Gawish, Fathyia; El-Khayat, Hanaa; Abu El Einin, Hanaa DREM was used to investigate the regulation of yeast metabolism Theodor Bilharz Research Institute, Environmental Research and to limitation, followed by abundance of two major nutritional Medical Malacology, Cairo, Egypt sources, glucose as the primary carbon source and ammonium as the primary nitrogen source. Objective: Allozyme analysis was used for analyzing the Results: Transcriptome profiles for two different sets of mating system of Biomphalaria alexandrina as parents and their nutritional conditions were investigated for bifurcation points, progenies produced by self and cross fertilization through two first, separately, to investigate individual responses, and second, successive generations under laboratory conditions. as a single complete dataset to reveal information on general Results: The analysis detected 14 genetic loci in 3 enzymes; nutritional stress response. Then a subset of transcription profiles ACP (3 loci), LDH (5 loci) and EST (6 loci); twelve of them were belonging to transcription factors only was considered in a similar polymorphic (85.5%). The mean number of Allele (A) and average manner. Differences in “regulation-only” sub-network and the heterozygosity over Loci (H) showed higher genetic heterogeneity complete transcriptome set were pointed out and the two sets in parents and their 1st generation than that observed in the 2nd of data were used suggest co-regulation of gene expression generation with abrupt change of either increase or decrease as well as antagonistic behaviour of transcripts to contribute among progenies produced by self fertilization ( 1st or 2nd to identification of the regulatory cascade of events regarding generation). The dendrogram resulting from cluster analysis based nutritional stress in yeast. on the genetic distances suggests close relationships among Conclusions: It has been observed that bifurcation points differ progenies of 2nd generation (self or cross) that cluster together in response to abundance or limitation of various macro-nutrients at diverging points 0.187 and 0.272 forming distinct group. Then indicating differences in regulation of the yeast metabolism. parents cluster with their self progenies at diverging points 0.259 and 0.293 then with their cross progenies at diverging point 0.311 forming another group. Then the two distinct groups cluster together at diverging point. 0.353. Conclusions: The allozyme analysis under the effect mating system of B. alexandrina reveals that snails of the same generation have closer genetic distance. The genetic heterogeneity among snails decrease through generations and

74 ICSB 2008 snails becomes genetically closer. There is more expectation of pathway contributes little to arsenite’s toxicity. Furthermore, the mutations to occur in snails which practice self fertilization. model predicts that a significant amount of intracellular arsenite is protein-bound; thus, arsenite’s toxicity may be due to other forms DS1-1-55 of GSH consumption and the modification of protein thiol groups. We are currently developing experiments to validate the model’s Towards a genome-scale kinetic model of cellular assumptions and verify its predictions. metabolism Smallbone, Kieran; Simeonidis, Evangelos; Broomhead, David; DS1-1-57 Kell, Douglas B. University of Manchester, MCISB, Manchester, United Kingdom The interplay between glycolysis and the calcium signal and membrane potential of human neutrophils stimulated Objective: Two divergent modelling methodologies have been with FMLP adopted to increase our understanding of metabolism and its Brasen, Jens Christian; Olsen, Lars Folke regulation. Constraint-based modelling highlights the optimal path University of Southern Denmark, Department of Biochemistry and through a stoichiometric network within certain physicochemical Molecular Biology, Odense M, Denmark constraints. Such an approach requires minimal biological data to make quantitative inferences about network behaviour; however, Objective: Neutrophilic granulocytes are part of the innate constraint-based modelling is unable to give an insight into cellular immune system, which protects the host from pathogens substrate concentrations. In contrast, kinetic modelling aims such as bacteria and fungi. Neutrophilic granulocytes are to characterize fully the mechanics of each enzymatic reaction. dedicated to the one purpose of killing these pathogens and This approach suffers because parameterizing mechanistic when challenged with bacteria or bacterial compounds they models is both costly and time-consuming. We aim to develop instantaneously generate large amount of superoxide and other a methodology for building a kinetic model of a genome-scale reactive oxygen species (ROS), known as the respiratory burst metabolic network, based on limited existing information. (1). Besides the production of ROS the activation by bacterial Results: In order to create a genome-scale kinetic model (albeit compounds is also characterized by an activation of several signal of minimal complexity), three sets of information are required over transduction pathways and a massive calcium influx (2). The and above network stoichiometry: fluxes through the network, respiratory burst itself is limited by the ability of the cell to provide metabolite concentrations and elasticities (changes in reaction sufficient amounts of NADPH to the NADPH oxidase (1). The rates with reactant levels). Where available, these parameters interplay between all these phenomenona is considered in this are determined from existing kinetic models available from http:// experimental study. biomodels.net/. Where unavailable, parameters are estimated Results: A suspension of neutrophilic granulocytes is activated Posters using previously-defined methodologies such as flux balance with 1 µM FMLP and the responses are measured using different Dedicated analysis. Once all values have been determined, the kinetic model fluorescent and luminescent probes. We have found that is defined using a phenomenological (linlog) rate law, and applied inhibiting glycolysis affects the ability of neutrophils to respond to a recently developed genome-scale model of yeast kinetics. on a chemotactic stimulus. Both the membrane potential and the Conclusions: This is, to our knowledge, the most comprehensive calcium signal are affected. kinetic model of yeast metabolism created, incorporating all Conclusion: There is a strong coupling between glycolysis and available modelling knowledge. It represents the first step in an the membrane potential and the calcium signal. Together with the iterative process of defining metabolic kinetics at the whole-cell pentose phosphate pathway glycolysis is providing the reducing level, and may easily be updated as further kinetic parameters are power for the production of ROS, and these results show that measured. glycolysis also indirectly regulates the neutrophilic response. Acknowledgment: We thank Prof. Dr. med. Torben Barington for DS1-1-56 kindly providing the blood samples and for fruitful discussions and laboratory technician Anita Lunding for her assistance. Modeling the yeast response to arsenite stress 1. Weiss, S. J. 1989. Tissue destruction by neutrophils. In N Engl Garla, Vijay1; Tamas, Markus2; Schaber, Jörg3 J Med. 365--376. 1University of Oxford, Berlin, Germany; 2University of Gothenburg, 2. Tintinger, G. R., and R. Anderson. 2004. Counteracting Department of Cell and Molecular Biology, Gothenburg, Sweden; effects of NADPH oxidase and the Na+/Ca2+ exchanger on 3Max Planck Institute for Molecular Genetics, Berlin, Germany membrane repolarisation and store-operated uptake of Ca2+ by chemoattractant-activated human neutrophils. In Biochem Objective: Arsenic poses a serious threat to human health in Pharmacol. 2263--2271. many regions, but is also effective in the treatment of certain cancers and parasitic diseases. S. Cerevisiae serves as a model DS1-1-58 organism for the study of arsenic metabolism in eukaryotes. In S. Cerevisiae, Fps1p mediates passive Arsenite (As(III)) uptake; Systems biology of iron in the body - Modeling of body Acr3p controls As(III) export from cells; and Ycf1p sequesters distribution dynamics As(III)-Glutathione conjugates to the vacuole. Our objective da Silva Lopes, Tiago Jose; Reich, Jens Georg was to develop a model of arsenic metabolism that reproduces Max-Delbruck Center for Molecular Medicine, AG Bioinformatik, experimental observations and enables the simulation of different Berlin, Germany experimental conditions, allowing us to direct hypotheses for further experimental research. Objective: Iron in ionic form is absolutely indispensable for life, Results: Here, we present a mathematical model of As(III). The but is also a severe poison if accumulated. Iron depletion is one of model was implemented as a system of ordinary differential the most frequent pathological states - iron poisoning a frequent equations that describe the dynamics of the change in genetic defect. Balance between not enough and too much is concentration of involved proteins and metabolites. The model extremely difficult and is being maintained mainly by the liver, was tested by simulation of several different knockouts and the duodenum and the Reticulo Endothelial System (RES) as was consistent with experimental data. With this model, we can regulator instances. predict the distribution of various pools of intracellular As(III), We are at present developing systems biology of iron in the quantify the relative contribution of each pathway to As(III) human body. This comprises a theory of iron balance and detoxification, and investigate the effects of perturbations of regulation in each in individual cell as well as a global balance kinetic parameters on the dynamics of metabolite fluxes and As(III) and regulation of organs in the whole organism. Tiny amounts of signal transduction. radioactive iron do not disturb the balance state of the body, but Conclusions: Our model shows that glutathione depletion via may serve as empirical marker of the fluxes between cells and sequestration of arsenite-glutathione conjugates to the vacuole is organs, including uptake in the intestine and desquamation of limited; we therefore propose that GSH consumption through this body cells. This situation can be described by ordinary differential

ICSB 2008 75 equations. DS1-1-60 Results: A first stage has been reached: we present a flux balance model of iron in the whole body. It is mainly based on Overexpression NADH Kinase and its Impact on Xylose published isotope studies in numerous clinical settings from the metabolism in recombinant Sacchromyces cerevisiae last seven decades of research. The resulting flux model displays Hou, Jin1; Vemuri, Goutham2; Bao, Xiaoming3; Olsson, Lisbeth1 a pronounced flux and time hierarchy. It is adequate to mimic the 1Technical University of Denmark, Center for Microbial main empirical results obtained by iron dose into blood plasma, Biotechnology, Lyngby, Denmark; 2Chalmers University of stomach and as damaged red blood cells into the RES. Technology, Department of Chemical and Biological Engineering, Conclusion: The present state of experimental knowledge does Göteborg, Sweden; 3Shandong University, State Key Laboratory not fully specify the pertinent flux parameters; hence we had of Microbial Biotechnology, Jinan, China to characterize the totality of numerical values in the parameter space compatible with the empirical findings. Objective:The nucleotides NADH/NAD+ and NADPH/NADP+ play The next step of this project is the molecular level. Since we important roles in redox metabolism. In Sacchromyces cerevisiae, developed an understanding of the upper “layer” of regulation (the NADH/NAD+ and NADPH/NADP+ have separate and distinct organs), we believe that this will help to characterize the picture of roles in the metabolism and can not cross the mitochondrial cellular and molecular iron regulation. inner membrane, therefore a closed redox balance need to exit References: [1] Hentze MW, et al. - Balancing acts: molecular both in the cytosol and in the mitochondria. In this work, we control of mammalian iron metabolism. Cell. 2004;117(3):285-97. studied the effect of introducing an additional source of NADPH [2] Hengl S, et al. - Data-based identifiability analysis of non-linear and decreasing NADH accumulation in xylose metabolising dynamical models. Bioinformatics. 2007;23(19):2612-8 recombinant S. cerevisiae. NADH kinase were overexpressed [3] Nathanson MH, et al. - Iron absorption in normal and iron- either in the mitochondria or the cytosol by introduction the deficient beagle dogs: mucosal iron kinetics. Am J Physiol. 1985 POS5 gene or the POS5 gene lacking the mitochondrial targeting sequence into xylose utilizing S. cerevisiae strains. DS1-1-59 Results and Conclusions:The recombinant strains were analyzed in aerobic and anaerobic cultivations. The growth The role of translational regulation in Saccharomyces characteristics of the strain overexpressing NADH kinase in Dedicated cerevisiae galactose utilization mitochondria were only affected marginally, while the strain Posters van den Brink, Joost; Pronk, Jack T; Daran-Lapujade, Pascale; de expressing NADH kinase in cytosol showed different physiology.

Winde, Johannes H It produced more ethanol and less CO2. The xylose consumption Delft University of Technology, Department of Biotechnology, rate also increased due to the increased NADPH availability. Delft, Netherlands However, the reaction requires ATP, which is necessary for cell growth, the specific growth rate decreased distinctly. Objective: Efficient bio-ethanol production from complex feedstocks requires conversion of the small carbohydrate DS1-1-61 fractions. The concentration of galactose, one of the fermentable sugars for Saccharomyces cerevisiae, can vary from 1% to Initial changes in yeast glycolysis upon nitrogen starvation 3% of the total carbohydrates depending on the feedstock. are regulated by metabolism and protein degradation Galactose utilization in yeast has mainly been studied under together aerobic conditions and not under industrially relevant anaerobic, van Eunen, Karen; Bouwman, Jildau; Westerhoff, Hans V.; Bakker, fermentative conditions. The goal of this study is to investigate Barbara M. the control of galactose utilization in the transition from glucose to VU University, Molecular Cell Physiology, Amsterdam, Netherlands galactose catabolism under aerobic and anaerobic conditions. Results: In a batch fermentation with glucose-galactose Objective: Time-dependent regulation analysis allows studying mixture, galactose consumption was severely retarded under the changes in the regulation of the flux through the individual anaerobic conditions compared to aerobic conditions. A 5 h enzymes of the glycolytic and fermentative pathways. In this study delay was caused by a slower induction of the structural enzymes we analyzed the fermentative capacity of baker’s yeast cultivated of the Leloir pathway, which determined the initial galactose under glucose-limited conditions upon nitrogen starvation. consumption rate. We decided to study the differences involved During nitrogen starvation unspecific bulk degradation of proteins in inducing galactose catabolism in more detail by shifting a well- (autophagy) is occurring. Therefore, we expected a simultaneous controlled, glucose-limited chemostat culture to galactose-excess and proportional decrease of the activities of the enzymes, conditions. Aerobically, galactose was consumed after a 1 h which should lead to metabolite homeostasis. This scenario lag-phase, where the induction of the Leloir enzymes determined would exclude metabolic regulation. To see if autophagy would the galactose consumption rate. Anaerobically, galactose was be dominant initially, we have examined the first four hours of not consumed at all, due to inhibition of translating the Leloir nitrogen starvation in detail. enzymes. While aerobically the ATP level was stable, the ATP level Results: The results of time-dependent regulation analysis dramatically decreased under anaerobic conditions. showed diverse categories of regulation. Indeed some enzymes Conclusions: The slow induction of the galactose consumption were regulated more by enzyme capacity during this nitrogen rate in an anaerobic batch with glucose-galactose mixture was starvation period, corresponding our hypothesis. However, other determined by the translation rate of the Leloir enzymes under enzymes were rather regulated metabolically in the first hours, control of the GAL system. The translational inhibition after and shifted only later towards regulation via enzyme capacity. sudden glucose-depletion is most likely determined by a quick Conclusion: Our results and analysis have shown that not only decrease in energetic state under anaerobic conditions. Our autophagy but also metabolic factors are responsible for the study clearly indicates the importance of translational regulation decreased flux through glycolysis during nitrogen starvation. The in efficient utilization of mixed-substrates bySaccharomyces role of the metabolic regulation is still unknown and therefore cerevisiae. More specifically, translation initiation and translation measurements of intracellular metabolites and regulators of rates are important determining factors under anaerobic glycolysis, i.e. ATP, fructose-2,6-bisphosphate and trehalose-6- conditions. phosphate, are necessary. The hierarchical part of the regulation is quite subtle and it seems that the regulation is distributed over the whole gene expression cascade. Therefore, a more detailed approach is needed by measuring synthesis and degradation rates of transcripts and proteins. Current work focuses on such in depth analysis of metabolism and gene expression.

76 ICSB 2008 DS1-1-62 several aspects of the project. We will analyse the metabolic and gene-expression dimensions by Regulation Analysis, to quantify Correlated genes with two novel genes implicated in the relative contributions of metabolism and gene expression to calcium metabolism the regulation of all ATP sensitive fluxes. We will also quantify the Park, Inju; Hong, Seong-Eui; Jin, Sora; Kwon, Juntae; Kim, Do contribution of adenine nucleotides to the metabolic regulation. Han; Cho, Chunghee Prior to this detailed analysis we must establish the network Gwangju Institute of Science and Technology, Department of Life topology with respect to ATP, i.e. which enzymes influence ATP Science, Gwangju, Republic of Korea and which are sensitive to ATP. To this end we first develop -together with the Manchester group- a simplified model of the Objective: Comprehensive understanding of the cardiac muscle main pathways involved in ATP synthesis and consumption in the molecular and physiological events requires identification based on existing data and flux balance analysis. Second, we of unknown genes expressed in this tissue. Previously, we are constructing a database which comprises known interactions analyzed the mouse cardiac muscle UniGene library containing of all enzymes with adenine nucleotides, to facilitate refinement 827 transcript clusters, and 19% of these genes are unknown. of the model and the required experimental work. In an iterative We systematically identified 14 authentic novel genes abundantly cycle of modelling and experimentation we will then work towards expressed in cardiac muscle through various in silico and in a complete model of ATP regulation. vitro methods. Among these genes, we focused on two novel genes related to calcium metabolism. To extend our systematic DS1-1-64 investigation into the two genes, we attempted to identify genes exhibiting transcriptional correlation with the novel genes Algorithm for in silicooptimisation of enzyme concentration Results: We established the clustering algorithms using the proportions Python script language based on the Pearson’s correlation Mozga, Ivars; Lagzdina, Laura; Stalidzans, Egils coefficient (PCC) values. Using 8 available cardiac or muscle Latvia University of Agriculture, Biosystems Group, Jelgava, Latvia microarray data from GEO, we discovered the groups of genes which exhibited similar expression profiles across the data sets. Objective: development of in silico optimisation algorithm and it’s This suggests that they are correlated and co-regulated genes application to increase bioethanol yield in yeast gycolysis. with the two novel genes under a variety of conditions. We Results: optimisation algorithm is developed for in silico obtained original findings that a numbers of the correlated genes metabolism optimisation procedure. Optimisation algorithm are associated with various intracellular transport processes. In consists from 8 steps including also one cycle. Algorithm is addition, we confirmed that the proteins encoded from the two orientated on determination of rational sequence of enzyme novel genes have calcium-binding activity through calcium-overlay concentration optimisation starting the process with enzymes Posters assay. that have the highest effect/cost ratio. Effect is estimated using Dedicated Conclusions: Our data suggest that the two novel calcium- metabolic control analysis (MCA) theory. Costs of enzyme related genes are implicated in transport mechanism. The genes concentration modification in this paper are assumed equal for all may share common upstream processes. Our inclusive data enzymes. Each consecutive inclusion of an enzyme into the set of establish a firm basis for future investigation into the cardiac gene optimised parameters is assessed from point of view of financial network and functions of these genes. benefit. Algorithm is demonstrated applying it for optimising of enzyme’s DS1-1-63 proportions for bioethanol production using Saccharomyces cerevisiae (bakers’ yeast) glycolysis model (Hynne, 2001). Microorganism systems biology: Energy and SBML standard model and COPASI software is used. Ratio Saccharomyces cerevisiae (MOSES) (Ethanol flow/Glucose uptake) is chosen as optimisation criterion Mensonides, Femke1; Kell, Douglas2; Kuchler, Karl3; Reuss, to maximise ethanol amount per one unit of glucose. Eight Matthias4; Ruoff, Peter5; Wölfl, Stefan6; Westerhoff, Hans2; Bakker, reactions (from totally 24 reactions in the model) are modified Barbara1 changing enzyme concentration. Optimisation of concentration 1VU University Amsterdam, Dept. of Microbial Cell Physiology, of all the 8 mentioned enzymes gives increase of ethanol yield Amsterdam, Netherlands; 2The University of Manchester, by 8,93 %. The highest optimisation potential was reached by Manchester Centre for Integrative Systems biology, Manchester, modifying concentrations of the first 4 most efficient enzymes United Kingdom; 3Medical University Vienna, Max F. Perutz increasing the yield by 8,53 % leaving just 0,4% increase for Laboratories, Vienna, Austria; 4University of Stuttgart, Institute the 4 enzymes with lower effect/cost ratio. Four optimisation of Biochemical Engineering, Stuttgart, Germany; 5University of methods of COPASI software are compared by their convergence Stavanger, Dept. of Mathematics and Natural Science, Stavanger, speed and stability: random search, evolution strategy, genetic Norway; 6University of Heidelberg, Dept. of Pharmacy and algorithms and evolutionary programming. The best convergence Molecular Biotechnology, Heidelberg, Germany speed and stability of solution was demonstrated by evolutionary programming. Objective: Systems biology comes in two flavours (bottom-up Conclusions: Metabolic Control Analysis theory and in silico and top-down) and at present these two hardly meet. The aim of modelling can be applied to determine a cost efficient sequence the SysMO project MOSES is to connect the two by developing of enzyme concentration modification enzyme ranging them by an approach that starts both from the edges and the nodes of the the level of their influence on the optimisation criterion. Application network. This project is a joint effort of six European universities. algorithm is developed. The idea is to follow a domino strategy, by a) extending systems biology along the pathways that regulate cell function, b) beginning with regulation through the most connective small molecules, and c) proceeding with regulation of major pathways synthesizing or degrading those molecules at high flux rates. Once the network of regulation through such a molecule has been mapped and quantitatively understood (modelled), regulation through the second best connected molecule is addressed. To demonstrate the principles of the approach, we have started with ATP (and ADP, AMP) as the most connected molecule and focus on the main pathways involved in catabolism, anabolism, and ATP consumption without driving growth (‘maintenance’). The system is analysed step by step and dimension (e.g. metabolism, transcription etc) by dimension. Results & conclusions: The Amsterdam group will contribute to

ICSB 2008 77 DS1-1-65 equation systems constraining the ratios of fluxes converging to a common metabolite are derived automatically from the model Quantification of yeast metabolism: of a metabolic network. The framework is applicable with all Glycolysis as a test case metabolic network topologies, 13C isotopomer measurement Weichart, Dieter1; Wishart, Jill2; Spasic, Irena2; Carroll, Kathleen2; techniques, substrates and substrate labelling patterns - also Messiha, Hanan2; Malys, Naglis2; Simeonidis, Vangelis2; when measurement data is too scarce to allow robust nonlinear Swainston, Neil2; Khan, Farid2; Dunn, Warwick2; Smallbone, fitting for obtaining a global flux distribution. The core of Kieran2; Mendes, Pedro2; Broomhead, David S.2; Gaskell, Simon FluxDirect constitutes of flow and independence analysis of J.2; McCarthy, John E.G.2; Oliver, Stephen G.2; Paton, Norman2; metabolic fragments and techniques for manipulating isotopomer Westerhoff, Hans V.2; Kell, Douglas B.2 measurements as vector spaces. 1The University of Manchester, Centre for Integrative Systems Conclusions: FluxDirect is able to automatically construct and Biology, Manchester, United Kingdom; 2The University of solve analytic flux ratio equations from 13C measurement data. Manchester, Manchester, United Kingdom The power of FluxDirect is demonstrated by computing previously unavailable flux ratios from mass spectrometry isotopomer Objective: The main goal of the Manchester Centre for measurements of protein bound amino acids for Bacillus subtilis Integrative Systems Biology (MCISB) is to bring together the on several different substrates. The computational techniques technologies and skills necessary for the development of behind FluxDirect also facilitate many rational experiment planning quantitative Systems Biology. These encompass a wide range of procedures, such as the selection of the most useful metabolites experimental, mathematical and computational activities. While to measure and in silico calculability analysis of flux ratios. the methods developed are intended to be generic, we seek to demonstrate their utility with a real system. To this end, the initial DS1-1-67 focus of research is on the metabolic network of baker’s yeast (S. cerevisiae), which is an ideal organism in which to explore The model of transferrin uptake by cell: a novel mode of techniques for the development of integrated models of important TFR2-mediated iron sequestration in oxidative stress cellular systems because the organism is highly amenable to Yevshin, Ivan1; Sharipov, Ruslan1; Shatalin, Yuriy2; Naumov, genetic manipulations and to high-throughput technologies. In Andrey2; Ermakov, Gennadiy2; Potselueva, Margarita2; Sukhomlin, Dedicated this organism, the glycolytic pathway was chosen first, with the Tatyana2 Posters aim of creating a quantitative description of this system. 1Institute of Systems Biology; Institute of Cytology and Genetics Results: For this purpose, a number of parameters and variables SB RAS, Novosibirsk, Russian Federation; 2Institute of Theoretical were measured. For each enzyme of interest, kinetic assays were and Experimental Biophysics RAS, Pushchino, Russian Federation developed to estimate affinity constants and turnover rates. The proteins were purified using MORF or TAP-tag constructs, and Objective. Iron is a vital ion supporting cell proliferation, the assays were run at a standardised pH of 6.5 to mimic in vivo metabolism and many cell-specific functions. Oxidative stress conditions. In addition, turbidostat cultures were sampled during (OS) and inflammation result in increase of levels of both non- steady-state growth to determine cellular protein and metabolite transferrin bound iron (NTBI) and transferrin (Tf)-bound iron concentrations by mass spectrometry. The experimental data are (TBI). TBI is not available for ROS and, thus, does not catalyze stored in appropriate repositories and annotated to facilitate their HO• production. Cells express Tf receptors, TfR1 and TfR2, dissemination using Web Services. to uptake TBI. Apparently, Tf level decreases in OS, although Conclusions: The parameters and variables determined were there is no proper explanation of the phenomenon to date. We compared to literature values, and the resulting new model is hypothesize that in OS uptake of diferric transferrin (Tf:Fe2) by the currently calibrated, validated and compared in its characteristics cells specialized to sequester and/or to transfer iron is enhanced with models published previously. In summary, we have been due to specific TfR2 properties. Recently, we have observed a able to establish a series of experimental and computational non-linear pattern of Tf level decrease in plasma of the rats with methodologies for implementation of quantitative systems biology, transplanted ascitic tumour [1]. In line with our hypothesis, TfR2 and have achieved their full integration at several levels, using the main function is to regulate extracellular iron content rather than example of the glycolytic pathway in yeast. just to be OS sensor, as it is currently accepted. The main goal of the work was to construct a model of iron and Tf uptake by the DS1-1-66 TfR2(+) cells in OS and to test the hypothesis about the role of TfR2 in iron regulation. FluxDirect - an analytic and systematic framework Results. The model of TfRs-mediated iron uptake was for estimating metabolic flux ratios from 13C tracer constructed in terms of chemical kinetics. The system of ordinary experiments differential equations was solved numerically using BioUML Rantanen, Ari1; Rousu, Juho1; Kleijn, Roelco2; Zamboni, Nicola2; workbench (http://www.biouml.org). The values of model Jouhten, Paula3; Maaheimo, Hannu3; Sauer, Uwe2; Ukkonen, parameters were obtained from available literature. Dynamics of Esko1 Tf:Fe2 in plasma was simulated on the basis of time course of 1University of Helsinki, Department of Computer Science, total iron experimentally obtained in OS. TfR2 dynamics resulted Helsinki, Finland; 2ETH Zurich, Zurich, Switzerland; 3VTT Technical from increase of Tf saturation was simulated. Time courses Research Centre of Finland, Espoo, Finland of extra- and intracellular Tf were obtained and fitted to the experimental kinetic curves. Objective: In vivo measurement of metabolic fluxes by 13C Conclusions.Suggested model illustrates that Tf concentration isotopomer tracer experiments is able to produce accurate may fall down due to biphasic dependence on Tf:Fe2 uptake by information about the distribution of metabolic fluxes. This TfR2(+) cells, increased Tf saturation resulted from increase of information can be harnessed in many ways, e.g. to unravel NTBI in OS further activates TfR2-mediated Tf:Fe2 uptake due targets and mechanism of metabolic regulation. Current to TfR2 stabilization. So, both TfR2-mediated iron and Tf uptake computational methods for estimating the metabolic fluxes from should be enhanced in OS. 13C isotopomer measurement data rely either on a manual References: derivation of analytic equations constraining the fluxes or on the 1.Yu.V. Shatalin et al. (2008) Differential change of ceruloplasmin numerical solution of highly nonlinear global system of isotopomer and transferrin in plasma and ascitic fluid of tumor bearer, Siberian balance equations. In the first approach, analytic equations concilium, 2:59-65. have to be tediously derived for each organism, substrate or labelling patter, while in the second approach, the global nature of an optimum solution is difficult to prove and comprehensive physiological measurements to augment the 13C isotopomer data are typically needed. Results: In our computational framework called FluxDirect linear

78 ICSB 2008 DS1-1-68 of K12 and K13 codon mutated cells indicate that a strong correlation exists between the flow of glucose carbons towards Oxygen dependence of metabolic flux distribution, either increased anaerobic glycolysis, and resistance to apoptosis metabolite concentrations and gene expression of (K12), or increased macromolecule synthesis, rapid proliferation, saccharomyces cerevisiae CEN.PK113-1A and increased sensitivity to apoptosis. These characteristic Jouhten, Paula1; Wiebe, Marilyn2; Rintala, Eija2; Huuskonen, metabolic adaptations may reveal specific targets for therapeutic Anne2; Tamminen, Anu2; Toivari, Mervi2; Ruohonen, Laura2; strategies. Penttilä, Merja2; Maaheimo, Hannu1 1VTT Technical Research Centre of Finland, Helsinki, Finland; 2VTT DS1-1-70 Technical Research Centre of Finland, Espoo, Finland Concept of metabolic networks as atom networks and Objective: While the physiology of Saccharomyces cerevisiae generation of metabolic networks by network expansion has been extensively studied in anaerobic and aerobic conditions Ohta, Jun and during respiro-fermentative growth in glucose excess, its Okayama University, Okayama, Japan physiology in oxygen-limited conditions is less well understood. Here we determined the intracellular flux distributions at five Background: Detailed experimental information of the entire different oxygen concentrations in glucose-limited cultures at a metabolism is accumulating rapidly through metabolomics dilution rate of 0.1 h-1 with 20.9%, 2.8%, 1.0%, 0.5% or 0.0% studies. Our knowledge of metabolic pathways is not complete 13 O2 in the inlet gas by combining fractional biosynthetic [U- C] in spite of its long history of study. In E. coli, part of genes have glucose labeling with MFA [1]. Since metabolite concentrations not been assigned to specific functions, yet. Discovery of new and gene expression are variables in the regulatory networks metabolic pathways may be possible through metabolomics which determine the intracellular fluxes, some of these were also studies. In this situation, informatics techniques for metabolic measured [2]. pathway prediction would be valuable. Results: While ethanol production and altered metabolite pools Concept and Method: First, this paper introduces the concept were observed already in 2.8% oxygen, only minor differences of metabolic networks as atom networks. Information of chemical in the flux distribution were observed, compared to fully aerobic structure and atomic tracing can be expressed in atom network conditions. structure of metabolic networks. Then, this paper introduces In 1.0% and 0.5% oxygen the severely restricted respiratory approaches where new metabolic pathways are explored on rates resulted in progressively reduced fluxes through the hypothetical metabolic networks generated to keep chemical TCA cycle and direction of major fluxes to the fermentative stoichiometry. Metabolic networks are generated by network pathway. However, only in 0.5% oxygen, was the biomass yield expansion from seed compounds via balanced reaction pattern in Posters Dedicated exceeded by the yields of ethanol and CO2 and respirative ATP the approaches 1 and via simple cleavage/formation of chemical generation still provided a substantial 25% of the ATP demand. bonds in the approaches 2,, respectively. An approach by In oxygen-limited conditions: 2.8%, 1.0% and 0.5% oxygen, Hatzimanikatis and his colleagues is re-defined in the approach 1, gene expression and metabolite pools were in general similar and using concept of “gmetabolic network as atom network” and the distinct from fully aerobic and anaerobic conditions. In anaerobic term “network expansion” introduced by Heinrich’s group. conditions an extensive redistribution of fluxes was observed Discussion: Pathway prediction would be an application of compared to all aerobic conditions. Positive correlation between generation of metabolic networks as atom networks by network the transcriptional levels of metabolic enzymes and corresponding expansion. Further, comparison of networks generated as above fluxes was found only in the respirative pathway. with real metabolic networks would contribute to understanding Conclusions: Although the cellular metabolism was respiro- of organization of metabolic system. fermentative in all the oxygen-limited conditions, the actual amount of oxygen available resulted in different flux distributions. Dedicated session 1-2: Fluxes, metabolites and genes responded to different external oxygen availabilities with distinct patterns. 1. Jouhten et al. BMC Standards and repositories Syst Biol (2008) 2:60. 2. Wiebe et al. FEMS Yeast Res (2008) 8:140-54. DS1-2-08

DS1-1-69 Integrative search engine and data warehouse for systems biology Comparison of metabolic flux distributions in NIH 3T3 Ramirez, Fidel; Albrecht, Mario mouse fibroblasts with K-ras codon-specific mutations Max Planck Institute for Informatics, Computational Biology and de Atauri, Pedro Ramón; Vizán, Pedro; Benito, Adrián; Selivanov, Applied Algorithmics, Saarbrücken, Germany Vitaly; Cascante, Marta University of Barcelona, Biochemistry and Molecular Biology, Objective: Nowadays it is recognized that the integration Barcelona, Spain of biological data is a critical step to achieve systems level understanding of complex cellular processes. This is reflected Objectives: Among K-ras mutations, codon 12 mutations have in the growing number of data warehouses for systems biology been identified as those conferring a more aggressive phenotype. that facilitate integrative data analysis by incorporating biological This aggressiveness is primarily associated with slow proliferation and medical information from different sources. The enormous but greatly increased resistance to apoptosis. Using transfected amounts of available data also necessitate efficient tools to find NIH 3T3 fibroblasts with a mutated K-ras minigene either at and retrieve relevant information. In particular, such tools must be codon 12 (K12) or at codon 13 (K13), the flux distributions able to cope with structured information as well as unstructured are determined in this two transfectant cell lines integrating text, which may hinder automatic processing. isotopomer distribution on metabolic products synthesized Results: We have built a comprehensive online resource that 13 from [1,2- C2]-glucose with software tools that can analyze the integrates information related to human genes and proteins from isotopomer accumulation in metabolites. over a dozen external databases. This resource currently includes Results: We show that codon 12 mutant K-ras (K12) transformed Gene Ontology annotations of human genes and proteins, cells exhibit greatly increased glycolysis with only a slight increase sequence family classifications, protein domain architectures, in activity along pathways that produce nucleic acid and lipid metabolic and signaling pathways, protein interactions and synthesis precursors in the oxidative branch of the pentose protein complexes, and disease associations. Our fast search phosphate pathway and via pyruvate dehydrogenase. K13 engine supports sophisticated queries for the integrated mutants display a large increase in pentose phosphate pathway information, and a well-designed web interface displays the activity and pyruvate dehydrogenase. results in a convenient user front-end. Our integrative warehouse Conclusions: The distinctive differences in metabolic profiles with three million annotations also enables systems biologists to

ICSB 2008 79 find the actual source of an interesting annotation quickly. Specific are mainly collected from the biological literature but we have also queries do not only return annotated biological concepts, but also complemented the collection with our own carefully designed compute relevant similarities between sets of genes or proteins problems. The collection includes both simple problems suitable using novel ranking algorithms. Furthermore, biologists can easily for initial testing, and more complex problems with realistic size track new updates of gene and protein annotations over time and quality of data. using RSS feeds. All problems are presented in a unified format and are fully defined Conclusions: The use of information retrieval methods in by a model space of allowed reactions (e.g. reaction kinetics like systems biology shows great advantages for uncovering Michaelis-Menten, or S-systems), ranges for the parameters, time important relationships between biological entities and series data that can be noisy, and an error function representing annotations that otherwise may remain hidden in the wealth the difference between simulated data from the model and the of data. A prototype of the search engine is available at http:// given data from the true unknown system. The objective is to biomyn.mpi-inf.mpg.de. minimize the error function, and hence ideally output the true system from which data originally was simulated. DS1-2-09 As initial benchmarks we report the results of our own identification algorithm (IET Syst Biol. 2007 1(2):120-9). For those The CellML model repository problems where the computation time of other methods has Lloyd, Catherine; Lawson, James; Hunter, Peter; Nielsen, Poul been published (about 15), this approach is currently orders of The University of Auckland, Auckland Bioengineering Institute, magnitude faster, and all problems have been solved without the Auckland, New Zealand use of supercomputing. Conclusion: A collection of benchmark problems for Objective: High throughput experimental techniques have lead identification of ODEs is presented. This enables objective to the population of web-accessible databases with vast amounts comparisons between different identification methods. of biological data, and mathematical models of biological systems are playing an essential role in the interpretation of this data. DS1-2-11 However, the scientific community now faces the challenge of the mathematical models themselves becoming increasingly complex Format overflow? Current developments and missing Dedicated and numerous, and there has arisen a need for centralised, elements Posters curated, online databases to store these models in standard Bittig, Arne T.; Schmidt, Henning formats and make them easily accessible and reusable. University of Rostock, Systems Biology and Bioinformatics Group, Results: The CellML Model Repository (http://www.cellml.org/ Rostock, Germany models) is such a database, providing free access to over 330 biological models derived from published, peer-reviewed papers. Objective: The development of the Systems Biology Markup These models describe a wide range of biological processes, Language (SBML) has been a milestone in the evolution including signal transduction pathways, metabolic pathways, of systems biology as a science. It enabled modelers and electrophysiology, immunology, the cell cycle, muscle contraction, researchers to easily share results, tool developers to use a and mechanical models and constitutive laws. This wide scope common format, and the emergence of model databases to exemplifies CellML’s ability to describe much of the biochemistry, store the results in a reusable format. However, in contrast to electrophysiology and mechanics of the intracellular environment. its name, SBML is not a format that encompasses the whole of Conclusions: CellML and the CellML Model Repository are systems biology, but merely a subset, only covering models that an integral part of the IUPS Physiome Project. As the CellML can either be defined by their biochemical components or freely community continues to grow, there will be more users submitting in terms of mathematical functions. Currently, more formats are their CellML models to the repository, and model curation and under development by the systems biology community. Examples annotation will become essential to the maintenance of the are MIASE, a format for storing Minimum Information About a CellML Model Repository as a useful resource. We anticipate the Simulation Experiment, and a still unnamed format for storing development and improvement of model simulation and editing data resulting from model simulations. However, while these two tools will further facilitate the model curation process, while model formats are of interest once a model has been developed, other annotation will be enhanced through the links with biological formats that would actually greatly support the model building ontologies. In addition, with the implementation of CellML 1.1, we process are unfortunately still missing. Two such needed formats intend to decompose the current CellML 1.0 models into a series comprise the formal representation of experimental settings and of reusable modules. As such the CellML repository will become experimental measurement data. An additional meta-format of a library of reusable models, facilitating the creation of new, more great use would be a format allowing the combination of models, complex models from pre-existing parts. experiment descriptions, and measurement data into modeling projects. DS1-2-10 Results: In this work we have developed a proof-of-concept implementation of experiment, measurement, and project A collection of benchmark problems for identification of formats. We demonstrate the advantage of their use in real ordinary differential equations modeling projects. Furthermore, we show that elements of Gennemark, Peter1; Wedelin, Dag2 the current MIASE specification can in principle be used to 1Göteborg University, Mathematical sciences, Göteborg, Sweden; represent projects. We point out missing features and motivate 2Chalmers University of Technology, Computer science and that, in contrast to what is often believed, it is possible to define engineering, Göteborg, Sweden a standard format for the most commonly employed types of measurement data in systems biology, which are important for Objective: We consider the problem of identifying both the modeling and parameter estimation purposes. structure and the parameters of an ordinary differential equation Conclusions: SBML is just for models. We need a similar (ODE) model from experimental data. In recent years, there has standard for experimental data and experiment descriptions. been a dramatic increase in the number of reported methods Ultimately it will allow the representation of modeling projects, approaching this problem in the biological literature. The natural speed up the modeling process and allow for the exchange of way to evaluate the performance of such methods is to try whole projects between different tools. them on a number of realistic test cases. However, lack of well defined and commonly accepted benchmark problems makes it difficult to judge the advantages and disadvantages of individual methods. Results: To enable objective comparisons between different methods we therefore present a collection of approximately 30 benchmark problems for ODE model identification. The problems

80 ICSB 2008 DS1-2-12 growing. In 2007, BioModels Database was ranked first data Using structured storage of π calculus models to enhance resource for Systems Biology [2]. And deposition of models upon re-usability and query publication is currently supported by the Nature Publishing Group, John, Mathias; Köhn, Dagmar; Heuer, Andreas PLoS and BioMed Central. University of Rostock, Computer Science, Rostock, Germany References: Objective: In the field of modeling biomolecular systems, the [1] N. Le Novère et al. BioModels Database: a free, centralized interest in the stochastic π calculus (C. Priami, Stochastic π database of curated, published, quantitative kinetic models Caluclus, 1995) and corresponding modeling and simulation tools of biochemical and cellular systems. Nucleic Acids Res, increases. Thus, the need for a common storage and exchange 34(Database issue):D689–D691, Jan 2006. format is given. The objective of this work is to introduce a [2] E. Klipp et al. Systems biology standardsthe community repository for stochastic π calculus models which provides speaks. Nature Biotechnology, 25:390–391, 2007. doi:10.1038/ features for simple but also for more sophisticated retrieval and nbt0407-390. query. Therefore, we introduce a common storage format that covers the major parts of the stochastic π calculus syntax and DS1-2-14 enables exact matching techniques. To extend query functionality, we investigate possibilities to apply different similarity measures, Framework for visualizing CellML Models coming on one hand from the field of model retrieval and on the Dissanayake, Sarala; Halstead, Matt; Nielsen, Poul; Lloyd, other hand from language theory. Catherine Our current approach of storing stochastic π calculus models The University of Auckland, Auckland Bioengineering Institute, uses an XML-based format called πML (D. Köhn, M. John, Auckland, New Zealand The π Markup Language, 2007). Following the ideas in (N. Le Novère et al., Minimum Information Requested in the Annotation Objective: CellML is an implementation-independent simulation of Biochemical Models (MIRIAM), 2005), we apply MIRIAM modelling language which is mainly used for understanding the guidelines to formal models. For example, process and channel dynamics of complex biological processes. Even though the names, that represent species and reactions respectively, CellML model structure and mathematics provide a powerful are annotated using common ontologies. For the querying of method for describing the dynamics of biological processes, models, there exist diverse matching techniques that enable the language has limited support for capturing higher level simple search, which can be applied to πML. However, more biological information. The biological knowledge is implicit in sophisticated model retrieval requires the introduction of similarity the mathematics and structure of the model. The objective is to measures. Therefore, methods for model comparison coming develop a framework for annotating and visualising biological and from the field of database research, e.g. model comparison biophysical concepts covered in CellML models. Posters based on ranking mechanisms, are used. Furthermore, the Results: A solution has been developed that involves: Dedicated applicability of formal approaches provided by language theory, • an ontology for storing CellML models in OWL format e.g. bisimulation, is discussed. How to integrate the latter into our (CellMLOWL) and a set of rules for binding these CellMLOWL model management concept, is subject to future work. instances to CellML models; Conclusions: The structured storage of stochastic π calculus • an ontology which represents biophysical and biological models in a common format does not only enhance querying, concepts that are covered in CellML models (CellMLBio) and a but also supports ideas like model exchange, model re-use, and set of rules for binding these concepts to CellMLOWL models; model retrieval. By combining database technologies with formal • a specification for building visual templates that support a methods, the search for stochastic π calculus models can not visual language that can be used to represent all biophysical only be performed on naming and ontology references but also and biological processes covered in CellMLBio ontology and on similarity matrices. the rules for binding the visual language to concepts within the CellMLBio ontology; DS1-2-13 • a visual editing tool, that combines the visual language and ontologies to visualize CellML models. BioModels Database, enhanced curated and annotated Conclusions: It is possible to provide a method of annotation resource of published quantitative kinetic models which allows the biophysics and biology of a CellML model to be Li, Chen; Le Novere, Nicolas represented and used for tasks such as visualization. EMBL-EBI, Computational Neurobiology Group, Cambridge, United Kingdom DS1-2-15

Objective: Quantitative models of biochemical and cellular Managing data from high content live cell microscopy systems are frequently applied to answer a variety of questions experiments in the biological sciences. The accessibility and quality of these Jameson, Daniel1; Swainston, Neil1; Spiller, David2; White, models is essential for their reuse. Michael2; Kell, Douglas1; Paton, Norman1 Results: BioModels Database (http://www.ebi.ac.uk/biomodels/) 1University of Manchester, Manchester, United Kingdom; is a free resource for storing, viewing and retrieving published, 2University of Liverpool, Liverpool, United Kingdom peer-reviewed quantitative models of biochemical and cellular systems [1]. Models are thoroughly curated and the model Objective: High content screening of microscopy images enables elements are annotated with terms from controlled vocabularies rapid identification of cell phenotypes across a wide variety of and links to relevant data resources. treatments. Such experiments generate large primary data sets A search engine in BioModels Database allows the user to with associated results that cannot be written into lab books. find models of interest. Alternatively, the user can browse a These may be stored on computer hard drives or DVDs where models-tree built based on Gene Ontology. The model can be typically only the experimentalist understands the precise system downloaded in various formats, which are SBML, CellML, BioPAX, of nomenclature that they have used for their generated files, and SciLab, XPP and VCell. Or, it can be simulated with an embedded experience suggests that even they may have some difficulty in simulation tool. Reactions network graphs of each model are discovering and interpreting datasets produced over an extended available in both SVG and GIF. A submodel of any model can be period. generated according to the users selection via an online tool. Presently our lab is performing in the region of 120 experiments As a supplement, the Web Services of BioModels Database per week, each one of which may be analysing up to 20 individual allows third-party software to directly access up-to-date data of locations containing multiple cells over variable time courses. models. We have sought to develop an information management solution Currently, models are accepted in both SBML or CellML format which as well as recording detailed metadata associated with the when submitted to BioModels Database. experiments, facilitates the correlation of results across imaging Conclusions: BioModels Database community is rapidly experiments.

ICSB 2008 81 Results: We have produced a data repository which automates (inside-outside functions), and polygonal surfaces. A series of some aspects of experimental data analysis and allows interaction compartment-geometric domain mappings were identified with with experimental metadata and results via a simple web different topology established. Spatial boundary conditions have interface. been introduced as “Rules” that are analogous to SBML initial Conclusions: Our repository is a useful and fully functional assignment rules. These extensions were prototyped using an example of how these data may be effectively indexed and internal SBML Java library adapted to VCell and exercised with managed to address the requirements of the end users. It existing VCell models. Although the work presented here focuses benefits from minimising the effort needed to input metadata and on continuous modeling with partial differential equations, the by adding value to the data in the form of automated annotation formal SBML Level 3 extension proposal will solicit contributions of results. from other groups involved in both continuous and stochastic spatial modeling. VCell is available at http://vcell.org and is DS1-2-16 supported by NIH P41-RR013186. This work is supported by NIH R01-GM070923. 4DXpress: Comparing gene expression data across species Haudry, Yannick1; Berube, Hugo2; van Noort, Vera1; Arendt, DS1-2-18 Detlev1; Bork, Peer1; Brazma, Alvis2; Furlong, Eileen1; Wittbrodt, Joachim1; Henrich, Thorsten1 Towards unifying systems biology - using pathway data in 1EMBL, Heidelberg, Germany; 2EMBL-EBI, Cambridge, United BioPAX format for SBML simulators Kingdom Blinov, Michael; Ruebenacker, Oliver; Moraru, Ion University of Connecticut Health Center, Center for Cell Analysis Objective: The increasing amount of data generated by high- and Modeling, Farmington, United States throughput technologies requires the integration of heterogeneous data and the development of tools in order to mine data, infer Objective: Thousands of biochemical interactions are available networks and extract knowledge. Databases like Ensembl have from public sources in the Biological Pathways Exchange demonstrated the invaluable benefits of data integration, from (BioPAX) format. However, the current standard for exchange of multiple resources, at the genomic level; allowing researchers to simulation-ready biological models is System Biology Markup Dedicated investigate comparative genomics. Gene expression patterns, Language (SBML). This markup language is structurally and Posters acquired through whole-mount in-situ hybridization experiments, semantically different from BioPAX. Some conversion schemes provide detailed spatial information on gene expression over time exist, using annotations and based on simple one-to-one during embryonic development. Despite the availability of good mappings between SBML and BioPAX objects, which ignores resources for different individual model species, this information is semantic differences and therefore often leads to significant loss rarely used from a systems biology perspective, due to a lack of a of information or meaning. A comprehensive modeling framework central repository allowing such comparison across species. capable of representing the complex relationships between SBML Results: We have integrated expression patterns of four and BioPAX data is needed to take full advantage of existing important developmental model species into a central repository pathway data in kinetic modeling, thus integrating these two called 4DXpress (http://4dx.embl.de/4DXpress/). Our first aim formats by gluing them together. is to provide an easy and comprehensive way for scientists to Results: Here we describe such a framework that we are access, query and analyze gene expression patterns across developing as a part of the Virtual Cell (http://vcell.org) modeling species through a web application. Cross-species relationships and simulation environment. Systems Biology Linker (Sybil, have been established between genes (orthology) and between http://vcell.org/biopax) is a tool for analyzing and visualizing time windows (developmental stages). We are currently BioPAX data, and converting them to SBML. Based on the Jena developing methods for mapping anatomical ontology terms Semantic Web framework for Java, Sybil supports handling of between species to identify homologous structures; a bilateria generic RDF/OWL data (such as visualization and reasoning) as ontology is under development for that purpose. To be able to well as functions specific to handling SBML and BioPAX data. compare gene expression patterns, we developed and compared Sybil uses Systems Biology Pathway eXchange, called SBPAX, as different similarity measures. In particular we are exploiting a generic approach to integrate model-centric formats similar to ontology structure to calculate semantic similarity between SBML with pathway-centric formats similar to BioPAX. SBPAX is expression pattern annotations. an OWL-based schema that serves as a glue to integrate different Conclusions: We have set up a resource integrating gene data formats, despite semantic differences. Sybil offers various expression information in space and time and defined a similarity visualization modes showing reaction networks to varying degrees measure which allows us to study co-expression networks and of details, including displaying nodes for reactions only as well as search for conserved network patterns across species using displaying Petri nets consisting of reaction nodes and reaction orthology relationships. We are convinced that our tool will help to participants and catalysts. Sybil also allows collapsing and understand evolution of genetic networks during embryogenesis. exploding various parts of the network individually, for example exposing reaction participant nodes to show all their components. DS1-2-17 Conclusions: SBPAX provides a bridge between SBML and the Semantic Web world. Toward an SBML extension for spatial modeling Moraru, Ion; Schaff, James DS1-2-19 University of Connecticut Health Center, Center for Cell Analysis and Modeling, Farmington, CT, United States The BMOND database – further development with new ideology The Systems Biology Markup Language (SBML) has become Sharipov, Ruslan1; Yevshin, Ivan1; Tolstykh, Nikita2; Likhovidova, an effective representation for exchange of quantitative, Elena2; Kel, Alexander3; Kolpakov, Fedor2 compartmental biochemical models. However, SBML Level 2 1Institute of Systems Biology; Institute of Cytology and Genetics cannot represent models with spatially resolved cellular geometry, SB RAS, Novosibirsk, Russian Federation; 2Institute of Systems inhomogeneous molecular distributions, and spatially localized Biology; Design Technological Institute of Digital Techniques processes. VCell, a mature modeling and simulation environment SB RAS, Novosibirsk, Russian Federation; 3BIOBASE, GmbH, for both spatial and non-spatial models, provides an appropriate Wolfenbuettel, Germany integration platform for the development of SBML Level 3 extensions for spatial modeling. Some of the required extensions Objective: The BMOND (Biological MOdels aNd Diagrams, former include spatial localization of reactions using a compartment Biopath) database specializes in reconstruction of complex attribute, and an option to interpret the reaction kinetics as time biological systems on the base of experimental data using rate of change of local concentration. A geometric representation method of formal description. New ideology and methodology was created based on geometric primitives, analytic geometry has been developed and started to implement making BMOND

82 ICSB 2008 more comprehensive and flexible and keeping it user-friendly. condition. This type of approach will be useful to evaluate the Results: Now BMOND contains about 500 diagrams and functional mechanisms of anticancer drugs. models created on the base of 1590 articles, of which regulation of eukaryotic cell cycle and cancer (253 diagrams and 629 DS1-3-13 corresponding articles), NF-κB pathway and inflammation (92 diagrams and 278 articles), nucleosomal regulation of Revealing the systemic modes of drug action by drug gene expression (45 diagrams and 434 articles), and essential scopes hypertension (66 diagrams and 94 articles), iron metabolism and Schwartz, Jean-Marc1; Nacher, Jose C2 oxidative stress (39 diagrams and 88 articles) and apoptosis 1University of Manchester, Faculty of Life Sciences, Manchester, modelling (8 diagrams and 45 articles) represent the most part. United Kingdom; 2Future University, Department of Complex Implementation of conceptions of composite diagrams and Systems, Hakodate, Japan BioHUB cross-linking public available databases UniProt, ChEBI, Gene Ontology, and EnsEMBL has been progressing. Objective: Drug development has traditionally been target- BMOND uses respective entities from these databases to build based, but this approach fails to take into account the diagrams and models. Abilities of working with microarray data multifactorial nature of many diseases. It is now widely admitted (manipulations, storage, web access) and results of analysis were that new systems-based approaches need to be developed to also extended. Application of new version of BioUML - our main understand drug action in large integrative systems and to assist tool for database content creation - allows to make this process the drug development process. more convenient and use advantages of novel methodology. Results: We introduce a new concept for the systemic analysis BMOND is also integrated with Cyclonet (http://cyclonet.biouml. of drug action termed the drug scope. Scopes were previously org) - our specialized database on cell cycle regulation and introduced in metabolic networks based on an expansion microarray data. BMOND is a core data platform in two European process: starting from a set if seed compounds, reactions whose projects FP6 “Net2Drug” and FP7 “LipidomicNet”. substrates are available are iteratively added and the final network Conclusions: New strategy of BMOND development as well as is the scope of the seeds. We similarly define a drug scope as unique content obtained by manual annotation makes it one of the union of scopes resulting from the expansion of all metabolic the outstanding databases among observed. Novel methodology targets of that drug. We constructed the scopes of 276 drugs is the next step to the database of new generation - BMOND-2. from the DrugBank database having metabolic targets. As scopes Availability: BMOND database is available: online at http://bmond. depend on the topology of the network used in the expansion biouml.org, and as MySQL dump or a set of text files (both by process, we considered four different systems for each drug: a request). human vs. a reference metabolic network consisting of the union

Acknowledgements: FP6 grant 037590 “Net2Drug”, FP7 grant of all organisms, and a reversible vs. an irreversible network. Posters 090107 “LipidomicNet” and interdisciplinary project 46 of SB Conclusions: The drug scope essentially represents the Dedicated RAS. maximum area of possible action of a drug in a metabolic system. Our analysis of 276 drug scopes in four different systems showed Dedicated session 1-3: Drug discovery that their distribution is uneven and that they can be grouped into different clusters. Some are small scopes associated to localised drug action, while others are large scopes associated to DS1-3-12 potential widespread systemic action. These classes are relatively well conserved throughout the four systems. We constructed Effect of anticancer drugs to upper limit dosege of the cell a network of drug scopes where scopes are connected if cycle related genes in S. cerevisiase their Jaccard distance is smaller than a given threshold, and Yoshida, Yuki1; Moriya, Hisao2; Kaizu, Kazunari2; Makanae, Koji2; an analysis of communities in this network revealed similar Kitano, Hiroaki3 characteristics. Overlapping of communities analysed by the 1The Systems Biology Institute, Systems Biology, Cancer Institute Clique Percolation Method may reveal key multi-role drugs. Institute, Tokyo, Japan; 2Cancer Institute, Systems Biology, Tokyo, However these communities are not correlated to specific Japan; 3The Systems Biology Institute, Tokyo, Japan therapeutical categories, reflecting different systemic modes of drug action rather than local interactions or chemical properties. Background: We previously reported a genetic method named genetic Tug-Of-War (gTOW) to determine an upper limit of gene DS1-3-14 dosage that maintains cell growth. In this study, we applied the gTOW method to investigate the possibility that identify unknown Utility of new open resources for accessing and comparing drug targets in the intracellular signal network. Many anticancer chemical structures of systems biology tool compounds drugs show various side-effects, but it is difficult to predict the and drug candidates effects and identify the targets. There is the possibility to detect Southan, Christopher of the target molecule and signaling pathway by comprehensive ChrisDS Consulting, Göteborg, Sweden analysis of gTOW with drugs. Methods: For yeast experiments cultures were two-overnight in Objective: While the use of small-molecules perturbagens 96-well master plate with SC+Leu medium at 30C. The cultures is a fundamental experimental approach in systems biology were diluted into media and dispensed into 384-well assay plates comparing compound structures has hitherto required expensive (50 ul wells containing media and test compounds) to yield ~500 database subscriptions. However, the last few years have seen cells/well. The plates were incubated in the presence of drug or a revolution in public cheminformatic resources (Southan et media at 30C for 50 hr. Upon completion, cell population and al. PMID:17897036). This work evaluates what can already copy number of gTOW plasmid with GFP were estimated by be used by systems biologists, using as an example a recent measuring turbidity and fluorescence at a single time point every Nature review of key small molecule inhibitors of protein-protein 30 min. All the statistical analysis was done by two-tailed Student interactions (SMIPPIs) that have utility both as systems biology fs t-test. tools and therapeutic candidates (Wells etal. PMID:18075579) Results: We tested the effects of three anticancer drugs, such Results: Using several sources it was possible to convert as 5-fluorouracil, doxorubicin, methotrexate, and used 30 types most small-molecule code names in the review to NCBI of gene related to cell cycle for the gTOW analysis. The growth PubChem compound identifiers, e.g. the IL-2 antagonist Ro26- inhibition rate of methotrexate was significantly increased in the 4550 to CID:16760522 and the BCL-XL binder ABT-737 to presence of gTOW plasmid of CDC15 which is protein kinase of CID:11228183, Others such as the HDM2-binder Nutlin-3, could the mitotic exit network. Our data suggested that methotrexate also be mapped to the EBI ChEBI database as CHEBI:46742. affects the CDC15-dependent signaling pathway. We are trying These were then searched against the 20 million compounds in to improve the sensitivity and reproducibility of this method PubChem to establish their chemical similarity neighbours, links by investigating plasmid copy number controlled by culture to PDB structures, target sequence homology relationships,

ICSB 2008 83 publications and PubChem bioassay data. In the case of of NRPSs enables the combination of modules from different Ro26-4550 using the PubChem SMILES representation with origin, which can then lead to the production of novel peptides. the free IBM patented chemical search established a direct Recently, communication-mediating (COM) domains, which allow match to US6806279 filed by Sunesis Pharmaceuticals in selective interaction between modules on separated polypeptide 2001 and, via direct PubChem links, to their 2003 publication chains, have been identified. This project aims at establishing PMID:12656598. Searching EBI Patent Abstracts identified further yeast as a cell factory for production of existing as well as novel Sunesis patents including IL4 antagonists in US6376524. From NRPs. structures exemplified in the downloaded patent a free Optical Results: As a model NRPS, we have heterologously expressed Structure Recognition tool was used to covert the images to δ-(L-α-aminoadipyl)-L-cysteinyl-D-valine synthetase (ACVS) of chemical structure representations. While there was no exact P. chrysogenum in S. cerevisiae. Since NRPSs depend on a match in PubChem the IUPAC name had a Google hit from covalently bound 4´-phosphopatetheine cofactor for activity, freepatentsonline. three different 4’-phosphopantetheinyl transferase encoding Conclusions: Systems biologists can now use powerful yet genes were co-expressed with ACVS encoding pcbAB. In all simple open cheminformatic resources as a crucial adjunct to three cases, ACV production was proven by LC-MS analysis. In the literature, their own data analysis, compound acquisition addition, we have tested the applicability of COM domains for choices, and designing new experiments. This will be massively NRPS expression in yeast by expressing a module of the Bacillus enabling for both the systems biology and medicinal chemistry brevis tyrocidine gene cluster together with its native COM communities domain and a module of the Bacillus subtilis surfactin gene cluster containing a mutated COM domain from different plasmids. Yeast DS1-3-18 strains containing both plasmids are showing production of the expected dipeptide. Systems analysis reveals the differentiation mechanism Conclusions: These results clearly indicate that S. cerevisiae is a of leukemia cells to macrophages induced by Ganoderma suitable host for production of NRPs. lucidum polysaccharides Hsu, Jia-Wei1; Huang, Hsuan-Cheng2; Juan, Hsueh-Fen3 DS1-3-21 1National Taiwan University, Institute of Molecular and Cellular Dedicated Biology, Taipei, Taiwan; 2National Yang-Ming University, Institute The interaction of nanoparticles for drug delivery with Posters of Biomedical Informatics, Taipei, Taiwan; 3National Taiwan biomimetic model membranes University, Department of Life Science, Taipei, Taiwan Frost, Rickard; Kunze, Angelika; Edvardsson, Malin; Kasemo, Bengt; Svedhem, Sofia Objective: Polysaccharides extracted from Ganoderma lucidum Chalmers University of Technology, Department of Applied (G. lucidum) have long been used to modulate the immune Physics, Göteborg, Sweden system in preventing or treating various human diseases in Asian countries. In our previous study, we found that fucose-containing Drug delivery of biopharmaceutics to the body is in many polysaccharide fraction F3 extracted from G. lucidum can bring cases limited by the fragile nature of the active compounds. A about cytokine secretion and cell death in human leukemia THP-1 current trend to avoid rapid degradation of therapeutic proteins cells. The results prompted us to further investigate on how F3 and peptides when inserted into the body, is to conjugate the induces anti-tumor effects in human leukemia. biomolecules to nanoparticles . The particles can be designed not Results: In this study, we integrated time-course microarray only to protect the biomolecule but also to deliver it to its site of analysis and network modeling to study the F3-induced effects action. Common designs of drug-carrying nanoparticles involve on THP-1 cells. Dynamic gene expression profiles showed liposomes, polymeric materials or inorganic particles. that cell differentiation was induced in F3-treated THP-1 cells. Objective: In this study, we aim at investigating how Furthermore, F3-treated THP-1 cells exhibited enhanced different nanoparticles interact with cell membranes. Different macrophage differentiation, as demonstrated by changes in nanoparticles have been exposed to biomimetic surface- cell adherence, cell cycle arrest, nitroblue tetrazolium reduction, supported lipid bilayers, and their interactions have been analyzed and expression of differentiation markers including, CD14, with the Quartz Crystal Microbalance with Dissipation monitoring CD68, matrix metalloproteinase-9, and myeloperoxidase. In (QCM-D) technique. Complementary data by the optical method addition, caspase cleavage and p53 activation were found to reflectometry have also been obtained. be significantly enhanced. To verify our results, general caspase Results: The results show the role of electrostatics in the inhibitor and p53 inhibitor were used to show that CD11b and interaction of nanoparticles and lipid membranes (using differently CD14 expression was reduced in F3-treated THP-1 cells. charged model membranes), and also indicate that lipid exchange Conclusion: Our data indicated that F3 is able to induce occurs between liposomal nanoparticles and the membranes. differentiation of THP-1 cells into macrophages by activating Implications for the design of drug-carrying nanoparticles are the caspase cascade and p53. Our results provide a molecular discussed. explanation for the effects of F3 on human leukemia cells through Conclusions: Model lipid membranes and surface sensitive a systems biology approach and offer a prospect for potential techniques provide ways to characterize, in vitro, the interactions leukemia and cancer therapy. between nanoparticles and biomimetic membranes.

DS1-3-19 DS1-3-22

Production of non-ribosomal peptides in Saccharomyces A novel approach to screen for regulators of autophagic cerevisiae activity Siewers, Verena1; Huang, Le1; Chen, Xiao1; San-Bento, Rita1; Hundeshagen, Phillip1; Eils, Roland2; Brady, Nathan2 Nielsen, Jens2 1Universität Heidelberg - DKFZ, CellNetworks, Heidelberg, 1Technical University of Denmark, DTU-BioSys, Kgs. Lyngby, Germany; 2DKFZ, Theoretische Bioinformatik, Heidelberg, Denmark; 2Chalmers University of Technology, Department of Germany Chemical and Biological Engineering, Gothenburg, Sweden Objective: Autophagy is a process of self-digestion that Objective: Non-ribosomal peptides (NRPs) represent a large serves to maintain cellular homeostasis through bulk protein group of secondary metabolites many of which have antibiotic, and organelle removal via the lysosomal pathway. Defective immunosuppressive or antitumoral activities and they show a autophagy has been put in relation to a variety of diseases. remarkable structural diversity. NRPs are synthesised by NRP This includes neurodegenerative diseases (such as Alzheimer synthetases (NRPSs), large enzymes that exhibit a modular or Parkinson disease), where defective autophagy might result structure, in which one module usually catalyses the activation in the accumulation of toxic proteins. Tumorigenesis has been and incorporation of one amino acid. The modular structure also linked to autophagy, since regulation of autophagy involves

84 ICSB 2008 several tumor suppressor genes and is further connected Igor2; Faratian, Dana3; Langdon, Simon3; Mullen, Peter3; Harrison, to the regulation of apoptosis. Therefore the identification of David3 novel regulators of autophagic activity might not only help to 1University of Edinburgh, Centre for Systems Biology, Edinburgh, understand the role of autophagy in these diseases, but could United Kingdom; 2University of Edinburgh, Centre for Systems be further used to modify autophagic activity in order to support Biology, Edinburgh, United Kingdom; 3University of Edinburgh, treatment or therapy. Edinburgh Breakthrough Research Unit & Pathology, Edinburgh, Results: We have established a novel approach that allows United Kingdom high-throughput screening for regulators of autophagic activity. Previously used methods to detect autophagic activity Objective: To study the response of the ErbB2 signalling network required high-resolution imaging techniques and only allowed to external stimuli in the context of ovarian carcinogenesis and detection of autophagosome formation. In contrast, our cancer cell resistance to receptor tyrosine kinase (RTK) inhibitors. technique allows detection of both, formation and degradation Results: A dynamic model of ErbB2 signalling network has of autophagosomes, using high-throughput methods such as been developed and validated by experimentation. The model flow-cytometry, plate reader and automated microscopy. Using includes Raf/MEK/ ERK and PI3K/PTEN/Akt signalling pathways this novel approach, we have systematically screened small- and accounts for the effect of ErbB2 inhibitors. In our model we compound libraries for new regulators of autophagic activity. assumed two-stage mechanism of binding of the inhibitor with the Identified compounds have then been characterized with regard receptor, based on the available literature on the inhibition kinetics to their level of interference within the autophagic regulatory of RTK dimerisation. machinery. The model was applied to interrogate the role of key signalling Conclusions: Recent research has linked autophagy to several proteins in modulating sensitivity to the microenvironment and to diseases. Depending on the context, this might be due to either therapeutic inhibition of RTK in order to understand autonomy up- or down-regulated autophagic activity. Therefore appropriate of cancer cell growth and drug resistance mechanisms. The modification of autophagic activity might allow more effective results of the modelling reproduce satisfactorily the experimentally treatment of such diseases. We here present a novel approach observed profiles of ERK and Akt pathway responses to to systematically screen for chemicals that regulate autophagic heregulin-beta stimulation in ovarian carcinoma cell lines. We activity. In addition, newly identified regulators of autophagic also studied the role of ErbB2 overexpression in the therapeutic activity and their specific function in the regulatory networks of resistance to RTK inhibitors. autophagy and apoptosis will be presented. Conclusions: The experimentally observed differences in the kinetic response of ERK and Akt branches of signalling to DS1-3-23 heregulin-beta stimulation were explained in the model by novel

mechanisms of cross-talk between these two branches.The Posters The impact of receptor-mediated endocytosis on the results of modelling clearly demonstrate that the overexpression Dedicated clearance of therapeutic proteins of ErbB2 results in a loss of sensitivity of Akt signalling branch to Krippendorff, Ben-Fillippo1; Huisinga, Wilhelm2 RTK inhibitors, which may contribute significantly to the overall 1Hamilton Institute - National University of Ireland Maynooth, effect of therapeutic resistance. Hamilton Institute, Maynooth, Ireland; 2Hamilton Institute - National University of Ireland Maynooth, Maynooth, Ireland DS1-3-25

Objective: Protein-based therapy has become an important Drug Targetability estimation through comparative strategy that has greatly benefited people suffering from ‘pocket-omics’ of host and pathogen: A case study in various disorders, including cancer, Crohn’s disease, diabetes mycobacterium tuberculosis and multiple sclerosis. When analyzing clinical trails, protein Yeturu, Kalidas; Raman, Karthik; Chandra R, Nagasuma drugs often exhibit complex nonlinear pharmacokinetics. Indian Institute of Science, Bioinformatics Centre, Bangalore, India Nonlinear pharmacokinetics results in a lack of desirable dose proportionality and complicates the design of appropriate Although target identification has been well recognized as a dosing schemes. Elimination by receptor mediated endocytosis critical step in modern drug discovery, identifying the right (RME) controls uptake and clearance disposition processes of target is by no means simple. One of the main unanswered numerous peptide and protein drugs. We present a systems problems with most clinically used drugs is that many of them biology model of RME in order to study the impact of RME on the exhibit adverse drug effects due to additional interactions with pharmacokinetics of therapeutic proteins and the arising of the unintended host proteins. A systems perspective of the proteome nonlinear pharmacokinetics. in terms of the interaction profile is essential to understand the Results: We present a kinetic cellular model that incorporates the pharmacodynamic outcome of a drug. No systematic method most important processes of RME w.r.t. drug pharmacokinetics. is available at present to address this issue, necessitating Following model reduction, we show that the clearance of the development of novel approaches. drug by RME follows a Michaelis-Menten type kinetics under An ideal target should first be essential to the pathogen, and the reasonable assumption that the receptor system is in steady preferably also unique, but should not share similarity in its ability state. We explicitly compute the parameters of the Michaelis- to bind drug-like molecules with proteins from the host. The host Menten type kinetics, Vmax and Km, in terms of the kinetic and the pathogen genomes can be compared computationally constants of the receptor system. Describing RME as a Michaelis- at various levels of abstractions, such as through their gene or Menten kinetic allows us to investigate the contribution of single protein sequences, protein structures, biochemical function(s) and cellular processes involved in RME on the arising nonlinearity, and systems level interactions. Here, we report the development of a the impact of drug-characteristics and cell types on the overall methodology which enables comparing the host and pathogen effect. proteomes by identifying the pocketomes in them and their cross- Conclusion: Our kinetic analysis elucidates the influences of matching. We have developed novel algorithms ‘PocketDepth’ the cellular processes on the nonlinear pharmacokinetics. This and ‘PocketMatch’ to identify putative binding pockets in protein mechanistic point of view contributes to our understanding structures and to compare them with each other. We have applied of drug elimination by RME, and therefore helps to better this approach to the human and M. tuberculosis proteomes as a understand and predict the nonlinear pharmacokinetics of case study and identify those pockets and corresponding proteins therapeutic proteins. in the pathogen that are targetable by drugs. For this purpose, 3,500 pockets from 767 shortlisted proteins from M. tuberculosis DS1-3-24 were compared against 70,149 pockets from 15,830 human proteins, involving over 245 million comparisons carried out on A systems approach to studying therapeutic resistance in a massively parallel IBM BlueGene system. Proteins containing cancer similar pockets are then clustered, and a targetability index is Goltsov, Alexey1; Moodie, Stuart1; Lebedeva, Galina2; Goryanin, computed. Besides targetability, biological insights obtained

ICSB 2008 85 about the similarities of binding pockets in diverse proteins will ion channels. This information by itself is insufficient to infer APD also be presented. This study provides a basis for a rational and change however. systems-level approach to understand drug pharmacodynamics Results: We have taken a publicly available model of cardiac and further to use such knowledge in discovery of new and safer action potentials (Hund Rudy) and calibrated it against a series drugs for tuberculosis. of DMSO-control experiments from a voltage-sensitive dye based assays on isolated canine ventricular myocytes (VM). This DS1-3-26 calibration allows us to capture the variation in inter- and intra- experimental drift and create an ensemble of parameter sets, Detecting liver enlargement based on the gene expression differing only in the initial concentrations of the extracellular ions. of rat liver induced by the treatment with different We have then simulated the anticipated change in APD for a compounds series of compounds with known channel inhibition profiles and Schmid, Ramona1; Ittrich, Carina1; Fundel, Katrin1; Lämmle, compared this to experimentally derived estimates in order to Bärbel1; Zapatka, Marc2; Brors, Benedikt2; Eils, Roland2; Weith, assess the predictivity of the model. Andreas1; Quast, Karsten1 Conclusions: Using a mathematical model of cardiac action 1Boehringer-Ingelheim Pharma GmbH & Co. KG, Biberach/Riß, potentials, we are able to integrate potency estimates from a Germany; 2DKFZ, Heidelberg, Germany range of ion channels and assess the potential QT risk that a particular compound may pose. This risk assessment can be Objective: An animal model was used to investigate the common performed at an earlier stage in the drug pipeline and with much toxic effects of a group of compounds that are well known to higher throughput than an in vivo assay. induce liver enlargement. The aim was to identify cellular functions affected in a similar way by the different compounds by means of Dedicated session 1-4: Plant systems gene expression. Methods: Sprague-Dawley male rats were treated with seven different compounds in two dosages together with the respective DS1-4-09 vehicle control. Gene expression profiles from liver tissue homogenates were derived after 6, 24, and 168 (192/ 336) hours A statistical analysis of relationship between root growth Dedicated using Affymetrix Rat Genome 230 2.0 Array. Linear models traits and expression of an auxin-responsive gene to Posters with dosage and time points as factors were applied to identify unravel molecular mechanism of plagiotropism of lateral differentially expressed genes. All genes with an FDR-adjusted roots in Arabidopsis thaliana p-value of the likelihood ratio test below 0.05 and a maximum Matsuzaki, Jun; Watahiki, Masaaki K.; Yamamoto, Kotaro T. absolute log ratio of the treated versus control group over all time Hokkaido University, Faculty of Science, Sapporo, Japan points above 0.5 were selected. Based on the genes fulfilling these criteria, gene set enrichment analyses were performed Objective: Most part of plant body is composed of lateral for each compound separately using the topGo package (www. organs. Growth orientation of lateral organs is a major bioconductor.com). Additionally, the enrichment in affected KEGG determinant of architecture of a plant body and thus of growth pathways was investigated using the globaltest package. Since and survival. We analyze molecular mechanism of plagiotropism all compounds share the same phenotype meta analysis was in lateral roots of Arabidopsis, which temporary grow obliquely used to join the results of the single compounds by combining the relative to the gravity vector (plagiotropism) after the initiation and p-values using Fisher’s omnibus analysis. then grow toward gravity vector (orthotropism). In order to unravel Results and Conclusions: Among the top scorers of the molecular mechanism of plagiotropism, it is necessary to quantify significant KEGG pathways are the ABC transporters, PPAR root growth traits and to analyze their relationship with localized signalling pathway, ether lipid metabolism, and fatty acid expression of mRNAs or proteins. metabolism. Additionally, significant GO Terms like sterol metabolic Results: Lateral roots of the mutant hy5 (Oyama et al., 1997) process, xenobiotic metabolic process, response to toxin, show plagiotropism for longer duration than wild-type. An auxin- microsome, ER, and iron ion binding clearly indicate a connection responsive gene MSG2/IAA19 is preferentially expressed on the to xenobiotic metabolism or stress in general and could also lower side of the elongation zone of a root during gravitropism. hint at liver enlargement. Genes which most contributed to the We introduced the MSG2 promoter-GUS fusion to hy5 and significance of these findings include CYPs, UGTs and ABCs. analyzed its expression in the elongation zone. Relationship These are well characterized genes in connection with xenobiotic between the MSG2 expressions and the root growth traits was metabolism as well as liver enlargement. One gene, which has analyzed using generalized linear mixed model. The MSG2 not been reported to be directly involved in xenobiotic metabolism expression was reduced as roots grew. However, it was sustained or liver enlargement so far, is Phopholipase A2. This gene is in longer roots in hy5 than in wild-type. Furthermore, differential responsible for the high scoring of several GO terms and KEGG expression of MSG2 in the elongation zone was observed less pathways mentioned above. frequently in hy5 than in wild-type when lateral roots of similar inclination were compared. Taken together, the hy5 roots grew DS1-3-27 with uniform MSG2 expression to be longer than did wild-type roots. Systems biology in drug safety assessment: Use of a Conclusion: We found that HY5 repressed transcription of recalibrated hund-rudy model to predict the effect of novel MSG2 in the elongation zone of root tips and promoted its drug compounds on action potential duration differential expression in response to gravitropic stimuli. Thus, Davies, Mark1; Abi-Gerges, Najah2; Pollard, Chris2; Swinton, reduction and differential transcription of MSG2 may reflect Jonathan1 transition of lateral roots from plagiotropism to orthotropism. 1AstraZeneca, ASTL, Macclesfield, United Kingdom; Quantification of traits on an organ scale and statistical analysis 2AstraZeneca, Safety Assessment, Macclesfield, United Kingdom of their relationship with expression of marker genes may be an effective approach to unravel molecular mechanism. Objective: Drug-induced ion channel modulation in ventricular myocytes can prolong action potential duration, which is seen on the electrocardiogram (ECG) as a prolongation of the QT interval. This is associated with a cardiac arrhythmia and therefore QT interval duration is carefully monitored in animal and man. Therefore, all candidate drugs are screened in an in vivo model of QT duration. This is however a late-stage and low throughput assay. For this reason, and much earlier in the drug discovery path, AstraZeneca screens compounds using IonWorks™ technology for their effects in vitro on the key cardiac

86 ICSB 2008 DS1-4-10 DS1-4-12

Salinity stress effects on antioxidant enzymes and Interlocked three-loop model, a new model of the biochemical components in two sorghum cultivars Arabidopsis circadian clock based on both biological data Heidari, Mostafa and robustness University of Zabol, Agronomy and plant breeding, Zabol, Islamic Saithong, Treenut1; Stratford, Kevin2; Painter, Kevin J.3; Millar, Republic of Iran Andrew J.4 1University of Edinburgh, Biological Sciences, Edinburgh, To evaluate effects of different salinity levels on antioxidant United Kingdom; 2University of Edinburgh, James Clerk Maxwell enzymatic active (CAT, APX and GPX) and osmotic components Building, Edinburg, United Kingdom; 3Heriot-Watt University, (carbohydrate and proline) in two sorghum genotypes an School of Mathematical and Computer Sciences, Department experimental was conducted as completely randomized factorial of Mathematics, Edinburgh, United Kingdom; 4University of design with three replication in zabol university at 2007. The Edinburgh, Centre for Systems Biology at Edinburgh, Edinburgh, treatments were three levels of salinity 0, 100 and 200 mM NaCl United Kingdom and two sorghum genotypes (Payam and Sistan). Results showed by increasing salinity levels from 0 to 200 mM NaCl, the activities A number of hypotheses on the molecular mechanism of the of these three antioxidant enzymes were significantly increased Arabidopsis circadian clock have been proposed based on and the highest level of active was in 200 mM salinity treatments. experimental data. For the purpose of confining the number of In the between of these two cultivars, except CAT enzyme, hypotheses, a series of mathematical models of the Arabidopsis Payam cultivar had the highest level of APX and GPX enzyme circadian clock has been developed. These models have been active. In this experiment, salinity had significantly changed on built based only on the system behaviours (mRNA time-series two osmotic adjustment concentrations (Carbohydrate and expression) without much consideration for matching the system proline) and increased their concentration. The highest of proline properties (robustness to variations). and carbohydrate were showed in sistan and payam receptivity. Objective:As these properties are of equal importance for In this study we found positive corelationship between osmotic developing a representative model of the real system, we have components and antioxidant enzyme active. This results observed the properties of the existing models by performing showed in these two sorghum cultivars these two mechanism sensitivity analysis. help to salinity tolerance, but payam cultivar had the highest Results:Particularly, the latest published model, consisting of concentration of osmotic adjustment, then its dry matter was three loops, shows low robustness to parameter variation which decreased higher than sistan cultivar. conflicts with the natural properties of a circadian clock. In this

Key words: salinity, antioxidant-enzymes, biochemical study, we present a new model of the Arabidopsis circadian clock Posters component, sorghum called ‘interlocked three-loop’, which may satisfy key properties Dedicated of the system as well as its behaviours. The new model was DS1-4-11 extended from the three-loop model by adding a single link between PRR9/7-LHY/CCA1 and LHY/CCA1-TOC1 loops in an Multiscale approach yields new insight into interlocking fashion. This type of circuit was suggested by our photomorphogenesis sensitivity analysis of simple oscillators of varied structures. Fleck, Christian1; Rausenberger, Julia1; Timmer, Jens1; Schaefer, Conclusions:The interlocked three-loop model, a new rational Eberhardt2; Kircher, Stefan2; Hussong, Andrea2 model suggested by mathematical analysis on model properties, 1University of Freiburg, Physics, Freiburg, Germany; 2University of is not only expected to improve robustness of the current three- Freiburg, Biology, Freiburg, Germany loop model, but it may also describe emerging experimental data.

Plants have evolved a variety of sophisticated mechanisms DS1-4-13 to respond and adapt to exogenic factors in their natural environment. Multiple photoreceptors regulate the plant’s The action of the POLARIS gene on the crosstalk between development according to the spectral quality and light intensity. auxin, ethylene and cytokinin signalling in Arabidopsis: In a combined experimental and theoretical approach we gain modelling and experiments new insight into the phytochrome B controlled signal transduction Liu, Junli Liu; Lindsey, Keith system in Arabidopsis thaliana. By suggesting a multiscale Durham University, School of Biological and Biomedical Sciences, model we are able to connect the mesoscopic phytochrome B Durham, United Kingdom protein dynamics to the macroscopic response, the hypocotyl length. Our model allows to single out the physiologically active Objective: To develop a mathematical model for the action of phytochrome pool which would be otherwise difficult to identify. the POLARIS gene on the crosstalk between auxin, ethylene and We show that the macroscopic response is very sensitive to the cytokinin signalling in Arabidopsis Hormone signalling systems total phytochrome amount and verify this finding experimentally. coordinate and to use the model to develop new insights into Furthermore, using our model we estimate the relevant dynamic how PLS regulates auxin concentration by controlling the relative parameters of the phytochrome pathway. We challenge the model contribution of auxin transport and biosynthesis and by integrating by light-dark transition experiments and find excellent agreement auxin, ethylene and cytokinin signalling. between our theoretical predictions and the experimental results. Results and Conclusions: First, the possibilities for the Hence, our linear response model captures the main features of interactions between PLS protein and the ethylene receptor phytochrome B mediated photomorphogenesis in Arabidopsis. ETR1 are analysed, and candidate interactions that result in observed mutant phenotypes are identified. In addition, analysis of experimental data reveals that regulation of ethylene on PLS transcription has to be independent of auxin. Based on these analyses, a model is subsequently developed. Second, the model is used to make quantitative predictions and further experiments are implemented. Modelling and experimental results for the effect of the pls gene mutation on endogenous cytokinin concentration are in agreement. Moreover, analysis on the discrepancy of auxin concentration in pls between modelling and experimental results reveals an additional role of PLS in auxin biosynthesis. This PLS regulatory role is included in the model. The model is able to reproduce all mutants available. Moreover, the model develops new insights into how PLS regulates auxin concentration by controlling the relative contribution of auxin transport and

ICSB 2008 87 biosynthesis and by integrating auxin, ethylene and cytokinin which cells within a homogeneous population of epidermal cells signalling. Third, model analysis reveals that a bell-shaped dose- develop into trichomes. We aim to elucidate this mechanism by a response relationship between endogenous auxin and root length combination of image analysis and mathematical modeling. is established via PLS gene. Model analysis develops insights Results:Our research is based on experimental data that has into how exogenous hormones affect root growth rate. Although been generated in this interdisciplinary project. Many different exogenously applying auxin, cytokinin and ACC generally reduces mutants have been produced and their influence on the trichome root growth rate, the underlying mechanism may be different. pattern has been characterized. We observe and quantify the This work shows that a combined experimental and modelling trichome pattern arising in growing leaves and determine the analysis for the complex interactions among hormones in plants influence of the leaf border and the age of the leaf. This in depth opens doors to develop new insights into how the cell integrates understanding facilitates the mathematical modeling that we use hormonal signals. to describe the distribution of activating and inhibiting substances. Testing the predictions of simulations against the experimentally DS1-4-14 obtained distribution of the trichome initiation sites allows us to link our model tightly to up-to-date research in experimental Dynamic features in a hierarchy of Calvin cycle models biology. Nikoloski, Zoran1; Grimbs, Sergio2; Selbig, Joachim2 Conclusions:In summary, our combined theoretical and 1University of Potsdam, Institute for Biology and Biochemistry, experimental approach yields new biological insights. Potsdam-Golm, Germany; 2Max-Planck Institute of Molecular Plant Physiology, Potsdam-Golm, Germany DS1-4-16

Objective: Analysis of existing models of photosynthesis coupled Applications of the second order kinetics in mathematical with experimental results have not been conclusive about the modeling of photosystem II function possibility that the Calvin cycle can operate in multiple steady Jablonsky, Jiri; Lazar, Dusan states under fixed environmental conditions. Our goal is three- Palacky University, Faculty of Science, Department of fold: (1) to theoretically analyze the existing Calvin cycle models in Experimental Physics, Olomouc, Czech Republic the framework of Feinberg’s Chemical Reaction Network Theory, Dedicated (2) identify a hierarchy of Calvin cycle models based on their ability Objective: For description of the redox reactions occurring within Posters to exhibit multiple steady states, and (3) to determine the key protein-complexes, the first order mass action theory should mechanisms—simplifications or augmentations—of a model that be used. However, using of the second order kinetics instead may lead to multiple steady states. of the first order kinetics significantly decreases the amount of Results: Our implementation of the Advanced Deficiency One model forms and therefore of differential equations of model. This algorithm overcomes the computational limits of the existing approach is questionable but sometimes may be justified. software solutions. None of the three studied models, (i) M. G. In the case of the photosystem II (PSII), the second order kinetics Poolman et al. Eur. J. Biochem. 268 (2001), (ii) G. Pettersson et should be used only for one reaction (electron transfer out of al. Eur. J. Biochem. 175 (1988), and (iii) X.-G. Zhu et al. Nonlinear PSII to the free plastoquinone pool). If the remaining main redox Analysis B (2008), satisfy the conditions of the Deficiency One reactions occurring in PSII monomer/dimer are described by the Theorem. Since these models violate the linearity conditions, first order kinetics, we obtain 576/1152 redox/model forms. investigation of nonlinear inequalities is required. All models To simplify models of PSII and to decrease their requirements for have a presignature, which may be used to confirm or deny the hardware, we have tested the usage of the second order kinetics existence of solutions for the nonlinear inequalities. in models of PSII. Conclusions: Simplification of model iii( ) by collapsing the Results: We developed three basic types of PSII models; reactions associated to Glyceraldehyde 3-phosphate results monomer of PSII with separated (model A; Jablonsky and Lazar in the simplest model of the Calvin cycle that exhibits multiple 2008, Biophys. J., 94:2725-2736) or connected (model B) steady states. Any other simplification by collapsing of reactions acceptor and donor side of PSII and dimer of PSII (model C) with leads to presignatures for which the linearity conditions are connected acceptor and donor sides of PSII. Electron transport violated. Augmentations of model (iii) by substituting the reactions between separated/connected acceptor and donor side(s) of PSII associated to Ribulose 5-phosphate, Ribulose 1,5-bisphosphate, was described by the second/first order kinetics. 3-Phosphoglyerage, and Glyceraldehyde 3-phosphate with We simulated the flash induced period-4-oscillations in oxygen those from model (i) do not result in pre-signatures for the linear evolution based on all three types of PSII models. The simulations inequalities. Therefore, caution is warranted in the level of detail based on models A and C quantitatively fitted experimental data included in the model, which can highly influence its ability for with almost same pattern of the oscillations. The model B was exhibiting multiple steady states. unable to describe experimental data because of zero oxygen yield after the second flash, i.e., simulated oscillation was without DS1-4-15 the so-called double hits. Conclusion: We show that it is possible to use simplified model New insights into pattern formation of trichomes on A (52 model/redox forms) instead of the complex model C (1152 Arabidopsis model/redox forms) for description of the oxygen evolution from Greese, Bettina1; Wester, Katja2; Bensch, Robert3; Ronneberger, PSII. Our results also partially support hypothetic cooperation Olaf3; Huelskamp, Martin2; Timmer, Jens4; Fleck, Christian4 of two PSIIs within dimer on the photolysis of water. However, 1Universitaet Freiburg, Physikalisches Institut / Fakultaet fuer more sophisticated analysis of the source of the double hits in the Biologie, Freiburg, Germany; 2Universitaet Koeln, Botanisches period-4-oscillations of oxygen evolution must be done to test this Institut III, Cologne, Germany; 3Universitaet Freiburg, Institut hypothesis. [Supported by grants MSM 6198959215 and 522/08/ fuer Informatik, Freiburg, Germany; 4Universitaet Freiburg, H003] Physikalisches Institut, Freiburg, Germany

Objective:Plants are beautiful examples of complex pattern forming events. An excellent model system for the study of cell differentiation and morphogenesis are the trichomes on leaves of Arabidopsis thaliana . Trichomes are hairs which are present on leaves, sepals and stems of many plants. They are regularly distributed over the epidermis, which makes them an ideal candidate for the study of spatial control of specific cell types in plants, since these structures arise by an orchestrated interplay of many different components. Some evidence has been accumulated that a cellular interaction mechanism determines

88 ICSB 2008 DS1-4-18 between infected and mock-inoculated leaves over time. The resulting network models reflect many of the regulatory Reconstruction and extension of the draft metabolic interactions known to occur during plant defence and provide a network of Chlamydomonas reinhardtii basis for the development of rational and experimentally testable May, Patrick1; Usadel, Björn1; Ebenhöh, Oliver2; Weckwerth, hypotheses. Wolfram2; Walther, Dirk1 Conclusions: Robust models inferred from the expression 1Max-Planck-Institute of Molecular Plant Physiology, Potsdam- profiles of differentially expressed transcription factors belonging Golm, Germany; 2GoFORSYS Institute of Biochemistry and to families with a known role in defence reveal the integrated Biology, Potsdam-Golm, Germany action of multiple WRKY proteins (plant specific transcription factors) and suggest a key role for the bZIP transcription factor Objective: There is an imparative need to integrate data from TGA3 in crosstalk between two key signalling pathways regulated high-throughput experimental techniques with bioinformatic and by salicylic acid and ethylene. theoretical modeling approaches to obtain a more comprehensive systems biology view of organisms. We believe an informative DS1-4-21 way to present an integrated view of the various levels of biological organization can be achieved via the visualization of The Integrative Screening Process (ISP) permits discovery the results within the biological context of metabolic pathways. of new treatment principles and prediction of therapeutic Therefore, we used metabolomics- and proteomics-assisted potential genome annotation to reconstruct the metabolic network of Kullingsjo, Johan; Andréasson, T.; Martin, P.; Pontén, H.; Chlamydomonas reinhardtii, a unicellular green alga. Sonesson, C.; Stansvik, A.; Svensson, P.; Waters, N.; Waters, S. Results: Metabolic pathway reconstruction was used to generate Neurosearch sweden AB, Pharmacology, Gothenburg, Sweden a pathway database for Chlamydomonas reinhardtii (ChlamyCyc), which currently features about 240 pathways with related genes, Objective: Most conventional drug discovery activities use enzymes, and compounds. ChlamyCyc was assembled from a sequential decision model, requiring a number of individual the available draft genome and our recently published MapMan criteria to be met. To a large extent the process is built upon in annotation using the Pathway Tools software within the BioCyc vitro screening on the molecular level. The assumption is made family of databases. The predicted pathways were verified using that one understands the disease mechanism(s) and thus, has orthology information from eleven other species. We analyzed extensive knowledge of the molecular target(s) for the treatment. and integrated a combination of database resources, such as As this is a questionable assumption within the Central Nervous existing genome annotations from JGI, EST collections, six- System (CNS), we have created as an alternative a systems frame translation of the genomic sequence, and protein domain biology drug discovery process, the Integrative Screening Process Posters scanning. Additonally, we used systematic high-resolution (ISP), directed towards creating novel treatments for brain Dedicated shotgun proteomics for correcting available or predicting new disorders. gene models as well as systematic metabolite profiling to project Method: ISP uses full-scale animal models and aims at mapping the identified metabolites onto the reconstructed metabolic draft the full response of the CNS to treatment. All new screening network. Furthermore, we applied structural modeling to the compounds are characterized using an extensive range of reconstructed metabolic network to identify minimum extension behavioral and neurochemical descriptors (brain response pathways on the basis of the presence of identified metabolites profiling). This process is optimized by utilizing a multivariate and proteins. modeling approach, including multivariate comparisons to a range Conclusions: ChlamyCyc and additional annotation data of clinical reference compounds and translation of biological is available under http://chlamycyc.mpimp-golm.mpg.de. effects and data from animal models into clinical effects in ChlamyCyc can be used as central reposistory for genomics humans. as well as for experimental data. It was engineered to visualize Results: ISP drug discovery has resulted in the discovery of new the Chlamydomonas functional molecular repertoire within the pharmacological principles, currently pursued in development context of metabolic pathways. towards new treatments. One example is the dopaminergic stabilizers, currently in Phase I-III clinical trials, demonstrating DS1-4-20 therapeutic benefit in schizophrenia, Parkinson’s and Huntington’s disease. Inference of transcriptional networks operating during Conclusions: We claim that the ISP provides an improved way of Botrytis cinerea infection of Arabidopsis discovering new treatment principles and predicting clinical effect Denby, Katherine1; Windram, Oliver P. F.2; Sharma, Sapna3; Wild, profile. The use of systems biology ISP has the potential to both David L.3 enhance the inventiveness and efficiency of the drug discovery 1University of Warwick, Warwick HRI and Warwick Systems process. Biology, Coventry, United Kingdom; 2University of Warwick, Affiliation: NeuroSearch Sweden AB, Arvid Wallgrens Backe 20, Warwick HRI, Coventry, United Kingdom; 3University of Warwick, SE-41346 GÖTEBORG, SWEDEN Warwick Systems Biology, Coventry, United Kingdom DS1-4-23 Objectives: We are studying the interaction between the economically important plant pathogen, Botrytis cinerea, and the Theoretical modelling reveals competitive complex model plant Arabidopsis thaliana. Recent advances in defence formation as the core of trichome patterning signalling research have revealed that different defence-related Geier, Florian1; Digiuni, Simona2; Schellmann, Swen2; Pesch, signalling pathways are interconnected in a network and that Martina2; Wester, Katja2; Dartan, Burcu2; Mach, Valerie2; Srinivas, extensive cross-communication occurs between pathways in Bhylahalli Purushott2; Hulskamp, Martin2; Greese, Bettina1; the network. Although numerous individual components of the Timmer, Jens1; Fleck, Christian1 plant defence mechanism have been identified, knowledge of 1University of Freiburg, Institute of Physics, Freiburg, Germany; the specific topology of the regulatory network is rudimentary. 2University of Cologne, Botany III, Cologne, Germany Our aim is to generate validated models of the transcriptional networks operating in the host plant after infection and to identify Objective: Trichome patterning in Arabidopsis serves as a model key regulatory hubs which affect susceptibility to this pathogen. system for de novo pattern formation in plants. It is thought to Results: We have used a variational Bayesian state space typify the theoretical activator-inhibitor mechanism although modelling approach developed in our group to reverse engineer this hypothesis has never been challenged by a combined transcriptional networks from high resolution, highly replicated experimental and theoretical approach. gene expression time series data obtained from Arabidopsis Results: By integrating the key genetic and molecular data of leaves infected with B. cinerea. Models have been built from the trichome patterning system we develop a new theoretical selected subsets of the genes exhibiting differential expression model which allows to directly test the effect of experimental

ICSB 2008 89 interventions and to predict patterning phenotypes. We show which acts as a transcription factor and is degraded after the experimentally that the trichome inhibitor TRIPTYCHON is interaction with phyB. transcriptionally activated by the known positive regulators Results: The PIF3 degradation and the effect of different light GLABRA1 and GLABRA3. Further, by particle bombardment treatments on this process are modeled by applying mass action of protein fusions with GFP we show that TRIPTYCHON and kinetics. We propose several molecular interactions leading to CAPRICE but not GLABRA1 and GLABRA3 can move between this degradation. Furthermore the non-equilibrium process of cells. Finally, theoretical considerations suggest promoter the clustering of phyB and PIF3 to nuclear speckles is described swapping and basal over-expression experiments by means of as a coagulation-fragmentation-process. Based on theoretical which we are able to discriminate three biologically meaningful considerations we suggest new experiments and find good variants of the trichome patterning model. agreement between the experimental data and the simulation Conclusion: Our work demonstrates that the mutual interplay results. between theory and experiment can reveal a new level of Conclusions: Experimental observations and biological understanding of how biochemical mechanisms can drive knowledge are integrated into a theoretical framework to biological patterning processes. understand the dynamics of the speckle formation and the involved interactions. Our combined experimental and theoretical DS1-4-24 approach provides a powerful tool to study the phytochrome B dynamics. A new technique for activation tagging in arabidopsis Pogorelko, Gennady DS1-4-26 NI Vavilov Institute of General Genetics RAS, Lab. of Functional Genomics, Moscow, Russian Federation The effect of sowing date and row spacing on yield and yield components on hashem chickpea variety under Objective: We have created and applied to Arabidopsis thaliana rainfed condition a new system of two vectors. The first vector (pEnLox) is intended Shamsi, Keyvan1; Mehrpanah, Hamid2; Kobraee, Sohil2; Nour- for activation tagging and contains a multimerized transcriptional Mohammadi, Ghorban3 enhancer from the cauliflower mosaic virus (CaMV) 35S gene in 1Islamic Azad University Kermanshah Branch-Iran, Plant Breeding Dedicated T-DNA flanked by two loxP-sites and the second vector (pCre) and Crop Production, Kermanshah, Islamic Republic of Iran; Posters contains the cre gene. Using pEnLox we have generated more 2Islamic Azad University Kermanshah Branch-Iran, Kermanshah, than a hundred mutants resistant to the herbicide ammonium Islamic Republic of Iran; 3Islamic Azad Univerity, Science and glufosinate, and about ten helper-lines resistant to the antibiotic Research Unit, Tehran, Islamic Republic of Iran hygromycin obtained with the use of pCre vector and also more than ten double mutants resistant to both selective markers. In Objective: In order to investigate the impacts of sowing date and at least 3 cases among 40 mutant lines that have been analyzed row spacing on yield and yield components of Hashem chickpea we observed constitutive ectopic expression of the genes veriety,a field experiment was conducted in 2005 at farm of adjacent to the T-DNA insertion that causes development of the Dorood Faraman(Kermanshah-Iran). In this study, the sowing mutant phenotype. Also, reversion of the mutants to the wild- date in three level (6, 23 November and 6 December) and the row type phenotype after removing the CaMV enhancer has been spacing in three level on rows (20, 30 and 40cm) were evaluated demonstrated. The system presented here provides a new and with complete randomized block design in factorial arrangement . easier way to analyze A. thaliana gain-of-function mutants. Results: Results of experiment showed that there are significant Results and Conclusions: At the current stage of development differences for planting date and planting density effects of plant of functional genomics in higher plants the “knock out” methods height, number of branch per plant; distance between 1st pod used to study gene functions do not allow exploration of the to soil, number of pod per plant, number of grain per plant, whole genome because of the presence of “silent” genes under biological yield and grain yield .The also maximum grain yield standard conditions of plant development. The currently available belong to sowing date 6 November and row spacing 30 cm. vectors designed to generate gain-of-function mutants are Conclusions: However maximum number of pod per plant not widely used due to inconveniences in the analysis of such and grain per plant belong to 40 cm row spacing but higher mutants. The unique system of vectors described in this work number of pod and grain per unit area in 20 cm row spacing will permit these problems to be avoided. The utility of the new result in increasing grain yield in this row spacing. As result system was demonstrated in three types of mutants. It allowed show increasing planting density resulted in decreased yield already at the initial stages 3 genes to be identified potentially components but from the other hand increased plant number affecting A. thaliana morphogenesis as a result of changes in their compensate decrease of yield components. Also we found that expression. planting at 6th December had higher distance of pod from soil surface and thus easier mechanized harvesting. DS1-4-25 Keywords:Chickpea,sowing date,Row spacing,Grain yield,Yield Components Modeling phytochrome B speckle formation and degradation of PIF3 DS1-4-27 Sonntag, Sebastian1; Kircher, Stefan2; Schaefer, Eberhard2; Timmer, Jens1; Fleck, Christian1 A miniclock in the unicellular picoeukaryote Ostreococcus 1University of Freiburg, Institut fuer Physik, Freiburg, Germany; tauri 2University of Freiburg, Institut fuer Biologie II/Botanik, Freiburg, Corellou, Florence1; Morant, Pierre-Emmanuel2; Schwartz, Germany Christian1; Thommen, Quentin2; Jacquet, Claire1; Motta, Jean- Paul1; Monnier, Annabelle3; Vandemoere, Constant2; Lefranc, Objective: Plants are sessile organisms and need to respond Marc2; Bouget, François-Yves1 and adapt to changes in their environment. To regulate the growth 1CNRS-Université Pierre et Marie Curie-Paris 6, UMR 7628 and development according to light intensity, spectral quality, Laboratoire Arago, Banyuls sur mer, France; 2CNRS-Université duration and direction of light, plants have evolved a system des Sciences et Technologies de Lille, UMR 8523 Laboratoire of various photoreceptors. The main photoreceptor mediating de Physique des Lasers, Villeneuve d’Ascq, France; 3Faculté de photomorphogenesis in red light is phytochrome B (phyB). Médecine, OUEST-genopole, Transcriptome Platform, IFR 140 A light-induced conformational change of phyB provides the GFAS, CS 34317, Rennes, France initial step in a signaling cascade finally leading to physiological responses. After photoconversion to its biologically active form, Circadian clocks result from the evolution of gene circuits phyB translocates to the nucleus, where it forms large protein based on interconnected feedback loops, which ensure the complexes, often called nuclear speckles. These early transient temporal coordination of biological processes among them speckles also contain the phytochrome interacting factor PIF3, and with external environmental cycles. We have developed

90 ICSB 2008 the unicellular green alga Ostreococcus tauri as a new minimal In this study we compare different models for the interaction model system to unravel the circadian clock architecture in this between CLV1 and RLK. We study how a present concentration very simple unicellular eukaryote with low gene redundancy. of CLV3 affects the expression of WUSCHEL, with and without By coupling experimental gene function analysis, based on interaction between CLV1 and RLK. genetic transformation and luciferase gene reporter strategy, to mathematical modeling we show that the only two identified DS1-4-31 clock genes, TOC1 and CCA1, assume central function in Ostreococcus in a reduced version of the circadian clock. Analysis and modeling of arabidopsis acid phosphatase- Quantitative agreement between experimental results and a encoding genes interactions minimal two-gene circuit model promotes Ostreococcus as a new Malboobi, Mohammad Ali1; Lohrasebi, Tahmineh1; Shojaie, model for system biology, with outstanding advantages such as Sharareh1; Manshaei, Roozbeh2; Sobhe Bidari, Pooya2; Feizi, a simple cellular organisation, compact genome with small gene Amir1 families and molecular tools for gene functional analysis. 1National Institute of Genetic Engineering and Biotechnology, Plant Biotech, Tehran, Islamic Republic of Iran; 2K.N. Toosi DS1-4-28 University of Technology., Medical Engineering, Tehran, Islamic Republic of Iran Genes, proline and total sugar accumulation in epidermis as well as in leaves of Arabidopsis under cold stress Objective: Phosphate anion (Pi) as biological important element Mohamed Suleiman, Khalifa1; Pearce, Roger S.2; Mohamed, is one of the least available plant nutrients in the soil. Acid Khalifa3; Pearce, Roger4 phosphatase (APase) enzymes play a key role in Pi acquisition 1University of Al-Jabal, Elgharbi, Libyan Arab Jamahiriya; from internal and external resources in all organisms. Unlike 2University of Newcastle upon Tyne, Newcastle, United Kingdom; animals, plants possess several acid phosphatases with broad 3Dr, Biology, Ghranin, Libyan Arab Jamahiriya; 4Dr, Biology, spectrums of substrates. The aim of this research is to predict the Newcastle opne Tyne, United Kingdom plant APase genes expression values and to model these genes biochemical network by artificial intelligent systems. The epidermis has an important role in plant physiology. Results: The design of gene-specific primers allowed expression How does it acclimate to cold? In this work, I compared analysis of each family member separately by quantitative- biochemistry and gene expression in the epidermis and whole comparative RT-PCR. The numerical data of gene expression leaf of Arabidopsis in the cold. The levels of sugar and proline in Arabidopsis thaliana plant produced in our lab were analyzed were similar in epidermis and whole leaves. Sugar and proline using Mamdani Neuro-Fuzzy Networks to predict gene accumulated when plants of Arabidopsis were exposed to expression values in Arabidopsis seedlings grown in different Posters cold. However, catalase, superoxide dismutase and ascorbate Pi concentrations. The proposed approach is based on finding Dedicated preoxidase activities were higher in leaves than epidermis in both a model to predict the expression pattern of a desired gene by control plants and in the cold. All the genes were expressed in analyzing other genes expression data in different concentrations epidermis as well as leaves, but the levels were mostly higher in of Pi. The achieved model provides a network by which we can leaves than epidermis in the cold. determine the rules of genes interactions in the class under study. The level of CBF1 was more highly expressed in epidermis at time Conclusions: Based on Mamdani Neuro-Fuzzy Networks, we 0 and 2 h, but in the leaves was expressed by 2 h in the cold. developed a software by which it is possible to obtain a model for However, CBF3 was earlier expressed in epidermis but was later predicting expression value of a specific gene and finding genes in leaves in the cold but, the level of COR15a was more highly interaction network. expressed in leaves than in epidermis. There were similarities and differences in the responses of leaves and epidermis. The DS1-4-32 pattern of CBF1 and CBF3 expression was different in different tissues and at different times in the cold. Total sugar and proline The clustering of arabidopsis acid phosphatases-encoding accumulated to the same level in epidermis as well as leaf but genes based on expression patterns activity of antioxidant enzymes was lower in epidermis than in Malboobi, Mohammad Ali1; Lohrasebi, Tahmineh1; Shojaie, leaves. Epidermis as well as leaf expressed nearly all genes tested Sharareh1; Sobhe Bidari, Pooya2; Manshaei, Roozbeh2; Faizi, in this work. Amir1; Malboobi, Mohammad Ali1; Alirezaie, Javad2 1National Institute of Genetic Engineering and Biotechnology, DS1-4-30 Plant Biotech, Tehran, Islamic Republic of Iran; 2K.N. Toosi University of Technology, Medical Engineering, Tehran, Islamic A study of the interaction between CLV3, its receptor CLV1, Republic of Iran and an additional unknown receptor-like kinase (RLK) in Arabidopsis thaliana Objective: Acid phosphatase (Apase) enzymes play a key role Sahlin, Patrik; Jönsson, Henrik in phosphate (Pi) acquisition from internal and external resources Lund University, Computational Biology and Biological Physics, in all organisms. Unlike animals, plants possess several acid Lund, Sweden phosphatases with broad spectrums of substrates. The aim of this research is to gain some insight to the plant APase The shoot apical meristem (SAM) is important for generation functional interconnection through gene expression analysis and of plant aerial organs. Undifferentiated stem cells reside within computational tools. the SAM and descendants are displaced to lateral positions for Results: The Design of gene-specific primers allowed expression differentiation. Different spatial domains of the SAM have different analysis of each family member separately by quantitative- gene expression patterns. Individual cells moves between these comparative RT-PCR. Expression analysis showed that the regions while the expression domains themselves stay constant. Apase-encoding genes could be clusered into four functional This robust behavior indicates that intrinsic intercellular signaling is groups: 1. highly expressed in Pi-starved shoots, 2. highly responsible for the spatial organization of the SAM. expressed in Pi starved in shoots and roots, 3. highly expressed In Arabidopsis thaliana the expression of the stem cell-promoting in Pi-fed in shoots and roots; and 4. highly expressed in Pi transcription factor WUSCHEL is controlled by the CLAVATA loci starved shoots and roots. In addition, the expression patterns (CLV1, CLV2, and CLV3). CLV3 is expressed in cells at the very tip of the Apase-encoding genes was different in various treatment of the apex. WUSCHEL is expressed in a region beneath the tip of duration times and tissues. The numerical data produced in our the apex. CLV3 binds to the receptor kinase CLV1 and initiates a lab and the microarray data through relevant available public signaling pathway leading to down-regulation of WUSCHEL. CLV3 databases were analyzed in a new approach named two-phase also binds to an unknown receptor-like kinase (RLK), suggested functional clustering too. The proposed approach is based on to have a functional overlap with CLV1. Mutations in the CLV loci finding patterns of time series gene expression data by Fuzzy can lead to ectopic accumulation of stem cells in the meristem. C-Means (FCM) and K-means clustering algorithms. First, genes

ICSB 2008 91 were clustered in each experimental condition individually and to discriminate between mutants in which a protein localizes then functional clustering based on all experimental conditions correctly, but exhibits a different temporal dynamics, or in was extracted using Pearson correlation between expression the case of lack/gain of function, when the spatio-temporal patterns of genes. A software was developed to visualize the dynamics is correct, but cell motility is impaired. Examples from interconnections among the APase-encoding genes from the Dictyostelium chemotaxis (Etzrodt et al., 2006) and mechanotaxis results obtained in the clustering stages. (Dalous et al., 2008) will be given. Conclusions: Plant Apases have different expression patterns Conclusions: So far QuimP has turned out to be a powerful in response to environmental conditions and in various tissues to tool to study cell motility. Future possible applications are nuclear fulfill cellular needs in a complex manner. Altogether, the compiled protein import/export and analysis of temporal gene expression data and the proposed methods allowed the visualization of patterns. possible interconnections among the APase-encoding genes. References: • J. Dalous, E. Burghardt, A. Müller-Taubenberger, F. Bruckert, DS1-4-33 G. Gerisch, T. Bretschneider. Reversal of cell polarity and actin- myosin cytoskeleton reorganization under mechanical and Growth and patterning of the arabidopsis leaf epidermis chemical stimulation. Biophys. J., 94(3):1063-1074, 2008. Robinson, Sarah1; Prusinkiewicz, Przemyslaw2; Coen, Enrico3 • M. Etzrodt, H. Ishikawa, J. Dalous, A. Müller-Taubenberger, 1John Innes Centre, Norwich, United Kingdom; 2University of T. Bretschneider, G. Gerisch. Time-resolved Front and Tail Calgary, CPSC, Calgary, Canada; 3John Innes Centre, Norwich, Responses to Chemoattractant in Dictyostelium cells. FEBS United Kingdom Letters, 580(28-29):6707-6713, 2006. URL: http://go.warwick.ac.uk/bretschneider Objective: The Arabidopsis leaf epidermis is a complicated two- dimensional tissue that consists of cells of various sizes, shapes DS2-1-10 and functions, which differentiate from equivalent cells. The aim of this project is to produce a quantitative model of a growing Noise from spatial heterogeneity changes the signal leaf epidermis that will help understand the relationships between amplification magnitude and increases the variability in growth, differentiation, cell division, and the final pattern of dose responses Dedicated differentiated cells. Of particular interest is the symplastic growth Kim, Jongrae1; Mao, Xuerong2; Heslop-Harrison, Pat3 Posters of epidermis, during which differentiated cells, some of which 1University of Glasgow, Glasgow, United Kingdom; 2University of unable to grow or divide, are accommodated into a growing Strathclyde, Glasgow, United Kingdom; 3University of Leicester, structure without sliding and tearing. Leicester, United Kingdom Results: In Arabidopsis, we can use a range of molecular markers to visualise cell structures and states of differentiation. Objectives: In most molecular level simulations, spatial We have developed the ability to image Arabidopsis seedlings heterogeneity is neglected by the well-mixed condition continuously for up to seven days using time-lapse confocal assumption. However, the signals of biomolecular networks microscopy. Using this technique we have followed the fate of are affected from both time and space. To account the individual cells, and extracted measurements about when, where spatial heterogeneity in the kinetic model without increasing and how they divide. Additionally growth measurements have computational burden exponentially, we consider multiple been made at the whole leaf, cellular and subcellular levels to help subvolumes of a reaction, introduce parameters representing understand the heterogeneous nature of the leaf epidermis and its transfer of ligands between the volumes, and reduce this to an coordinated growth. error-term representing the difference between the well-mixed Conclusion: Imaging a living tissue is a powerful way to condition and the actual spatial factors. The effect of varying determine what is happening before, during and after cell division, this term between 0 (well-mixed) and 1 (no mixing) and of not just in the dividing cell, but also in the surrounding tissue. adding noise to the kinetic constants was then investigated and Initial observations have made it possible to test existing theories correlated with knowledge of the behaviour of real systems and of cell division and in most cases resulted in their falsification situations where network models are inadequate. The spatial in the case of Arabidopsis leaves. The data obtained in our distribution effects on the epidermal growth factor receptor experiments can be used to construct a quantitative model of leaf (EGFR) in human mammary epithelial tissue are studied by growth and cell division. This model can also provide a framework introducing noisy kinetic constants. for understanding how complex patterns of differentiated cells are Results: The steady-state of the dose response in the EGFR accommodated into a growing tissue. is strongly affected by spatial fluctuations. The ligand-bound receptor is reduced up to 50% from the response without spatial Dedicated session 2-1: fluctuations and the variance of the steady-state is increased at least 2-fold from the one for no spatial fluctuations. On the other Cell-regulation - signalling hand, dynamic properties such as the rising time and overshoot are less sensitive to spatial noise. DS2-1-09 Conclusions: The EGFR signal transduction pathway may evolve to reduce the effect from the spatial heterogeneity - through active Mapping spatio-temporal dynamics of fluorescent proteins transport mechanisms, or selection to exploit rapidly transported in moving cells using the image analysis package QuimP ligands - or have some control architecture - localization of Bretschneider, Till reactions in appropriate sub-domains within the cell - to carefully University of Warwick, Warwick Systems Biology Centre, manage not only timing but also spatial distributions. From Coventry, United Kingdom the pharmaceutical point of view, since the dose response is significantly diminished by the spatial non-uniformity, to maximise Objective: The widespread availability of fluorescent fusion the response not only the amount but also the spatial distribution proteins in combination with high throughput methods to study of the dose must be carefully controlled - targeted to particular large numbers of single cells over time demands novel automated cellular domains, having active transport and passive diffusion analysis methods. To quantify patterns of cytoskeletal proteins in parameters appropriate for the reactions being manipulated. the cortex of moving cells, Dictyostelium or other amoeboid cells like neutrophils, we have developed QuimP software, a set of plugins for the image processing package ImageJ. Results: QuimP allows to follow local cortical fluorescence intensities in the reference frame of a moving cell and to correlate the fluorescence intensity with local protrusion and retraction activity of the cell membrane, as well as morphometric features. Such a more subtle analysis of mutants is possible, e.g. in order

92 ICSB 2008 DS2-1-11 type model with the Adair-Klotz framework widely used by experimentalists, we propose a way of relating observed apparent A Multi-step approach to dissect the gene network association constant to microscopic association constants for the regulated by the RNA binding protein Musashi1 T and R states. Penalva, Luiz O.1; Sanchez-Diaz, Patricia1; Burns, Suzanne1; Conclusions: We propose an allosteric model of calmodulin Burton, Tarea1; de Souza Abreu, Raquel1; Vogel, Christine2; Ko, that is consistent with experimental data and accounts for Daijin3 various known properties of calmodulin. Kinetic simulations 1UTHSCSA, San Antonio, United States; 2UT Austin, Austin, reproduce differential activation of PP2B and CaMKII at different United States; 3UTSA, San Antonio, United States calcium concentrations. The existence of open conformations of calmodulin in the absence of calcium is a testable prediction of Objectives: RNA binding proteins (RBPs) play an important our model. role in numerous biological processes and contribute to the final protein content of a cell in a quantitative and qualitative DS2-1-13 manner. Unfortunately, global studies covering functional and biological aspects of RBPs are rare. We conducted a systematic Mathematical analysis of EGFR signal transduction analysis of Musashi1 (Msi1), a RBP that bridges the fields of network associated with lung cancer stem cell biology and cancer. Its functions are required during Naruo, Yoshimi embryonic development while its aberrant expression has been Tokyo Medical and Dental University, Biomedical Science Ph.D. found in several types of tumor. In our studies we (1) carried out Program, Tokyo, Japan biological assays to demonstrate a direct link between Msi1 and tumorigenesis and (2) performed a multi-step expression analysis Objective: Ligand binding to receptors on the cell surface to identify the gene network regulated by Msi1. and activation of the associated proteins initiates a signaling Results: First, we knocked down (KD) the expression of Msi1 cascade that regulates various cellular processes. Epidermal in the medulloblastoma cell line Daoy. The effect of Msi1 upon growth factor receptor (EGFR) plays key roles in the regulation anchor independent growth was analyzed as an indicator of cell proliferation, differentiation, and survival in various tumors. of tumorigenesis. We observed a 3-fold reduction in colony Previous reports have shown that EGFR overexpression and formation when expression of Msi1 was depleted. In addition, an mutations in the EGFR tyrosine kinase domain are common in impaired ability to form spheres (hallmark of stem-cell phenotypic non-small-cell lung cancer (NSCLC), and these characteristics identification) was also observed in the KD; which suggests a role induce distinct responsiveness to EGF and EGFR tyrosine kinase for Msi1 in the maintenance of the stem-like tumor cells. Second, inhibitor. However, the detailed mechanism of the signaling to map the gene network regulated by Msi1, we employed three pathway directed transcription remains incompletely understood. Posters different methods. We first used Rip-Chip to identify the mRNA Results: Here, we used three different NSCLC cell lines; wild Dedicated population preferentially associated with Msi1. Second, we type parental cells, parental type cells with EGFR-overexpression, used mass spectrometry shotgun proteomics to determine the or L858R mutant EGFR-overexpression (EGFR kinase sensitive effects of Msi1 on protein expression. Finally, we performed a mutant) [1] to observe differential signaling kinetics and gene transcriptomic analysis using both microarrays and qRT-PCR. expression after addition of EGF or an EGFR inhibitor. Based Conclusion: We provide a detailed case study of an important on quantitative microarray and phosphoproteome datasets, we RBP. Our results represent the first experimental evidence identified expression of several cell-specific genes. Based on that Msi1 regulates a large number of genes and acts via very quantitative data with a newly identified regulator, we described a different routes. A substantial number of the target genes are network of EGFR signaling mathematically. involved in cell cycle regulation, apoptosis, cell proliferation and Conclusions: Our results demonstrated that EGFR cell differentiation. Msi1 also affects the expression of several overexpression and L858R mutation had influences on the genes connected to two crucial developmental pathways that are dynamics of EGFR signaling pathway. The result suggested that frequently altered in brain tumors: Notch and Sonic Hedgehog. kinetic variation, such as concentration and down-regulation rate of EGFR, might induce a difference in gene regulation [2]. DS2-1-12 References: [1] Chen, Y. R., Fu, Y. N., Lin, C. H., Yang, S. T., Hu, S. F., Chen, Y. T., Tsai, S. F., Distinctive activation patterns An allosteric model of calmodulin in constitutively active and gefitinib-sensitive EGFR mutants, Stefan, Melanie I; Edelstein, Stuart J; Le Novere, Nicolas Oncogene, 25(8): 1205-1215, 2006. EMBL-European Bioinformatics Institute, Cambridge, United [2] Nagashima, T., Shimodaira, H., Ide, K., Nakakuki, T., Tani, Kingdom Y., Takahashi, K., Yumoto, N. & Hatakeyama, M. Quantitative transcriptional control of ErbB receptor signaling undergoes Objectives: Calmodulin plays a vital role in mediating bidirectional graded to biphasic response for cell differentiation, J. Biol. Chem. synaptic plasticity by activating either calcium/calmodulin- 282(6): 4045-4056, 2007. dependent protein kinase II (CaMKII) or protein phosphatase 2B (PP2B) at different calcium concentrations. The aim of this study DS2-1-14 was to create a model that accounts for known properties of calmodulin that have been difficult to model so far. Integrated transcriptomic and proteomic profiling of Results: We designed an MWC model for calmodulin activation, multiple kinase inhibition: Towards an understanding of in which binding to calcium facilitates the transition between a drug action in tumour cells low-affinity (tense, T) and a high-affinity (relaxed, R) state. The four Nahnsen, Sven1; Rho, Seong-Hwan2; Schittek, Birgit3; Nordheim, calcium binding sites are assumed to be non-identical. The model Alfred4; Kohlbacher, Oliver1 comprises 352 reactions. Estimates for model parameters were 1University of Tuebingen, Center for Bioinformatics, Simulation of obtained from experimental calcium binding curves for wildtype Biological Systems, Tübingen, Germany; 2University of Freiburg, calmodulin, wildtype calmodulin in the presence and absence of Freiburg Center for Data Analysis and Modelling, Freiburg, several ligands, and mutant versions of calmodulin lacking either Germany; 3University of Tuebingen, Department for Dermatology, N-terminal or C-terminal binding sites. Kinetic simulations were Division for Dermatological Oncology, Tuebingen, Germany; then performed using COPASI. Simulation results are consistent 4University of Tuebingen, Institute for Cell Biology, Proteome with experimental datasets for calcium binding to calmodulin Center, Tuebingen, Germany that have not been used to constrain the model. Furthermore, the model reproduces experimental accounts of active sub- Objective: The human mitogen activated protein kinase and the saturated forms of calmodulin, an increase in calcium affinity phosphoinositide 3-kinase pathways are widely studied targets upon target binding and the differential activation of CaMKII and for cancer therapy due to their constitutive activation in many PP2B. We further predict the existence of open conformations human cancers. Multiple kinase inhibitors, such as Sorafenib in the absence of calcium. To facilitate comparison of an MWC- (BAY-43-9006) are auspicious drug candidates in cancer

ICSB 2008 93 therapy. In contrast to promising in vitro studies, the translational DS2-1-16 clinical investigations show that the inhibition of single pathways (monotherapy) is barely successful. We used transcriptomics and Prediction of transcriptional regulatory region and motif proteomics experiments to investigate the mechanism of multiple using support vector machine and principal component inhibition of signalling in tumour cells. analysis Results: mRNA profiling experiments of cancer cell lines revealed Ito, Masahiro; Takeda, Sachiko; Maruyama, Kozue; Ishido, the complex influence of drug mediated inhibition of signalling Yusuke; Yokoi, Kazuhito to gene expression. Our complementary proteomics approach Ritsumeikan University, Department of Bioinformatics and allowed for some candidate genes to confirm their involvement Biotechnology, Kusatsu, Japan in drug mechanisms, however a simple mapping of the gene expression profile to the next level of biological complexity, the Objective: Genes acquire functions through several regulations in proteome, was not possible for most of the genes. For some time-course and space-axis. The first gene regulatory regulation, proteins we observed high ratios of regulation, whereas the transcription, is the most important for determining whether to respective transcripts were regulated only slightly. For a high turn on or off the following regulatory process. Transcription is number of strongly regulated transcripts the corresponding an extremely complex process, but rational. The expression of translation products were either not regulated or not detected gene clusters with a series of functions, such as development, by our method. The pI and molecular weight range, imposed by differentiation and metabolism, is co-regulated by binding the DIGE (difference gel electrophoresis) methodology, allowed of transcription factors selectively to specific sequence sites to dissect genes that are technically not covered by 2D DIGE in the cis-regulatory regions of genes. Therefore, the genes from those that might fall under the abundance threshold of showing similar expression profiles measured by microarray concentration or whose regulation is biologically not ‘mapped’ to technology must be regulated by the same transcription factors the protein level. to the corresponding binding sites in regulatory regions. Some Conclusion: Our proteomics analysis showed mostly distinct prediction methods of regulatory regions were proposed. sets of regulated proteins upon drug mediated inhibition of major However, it was difficult to predict new regulatory regions, such as signaling pathways. This is in concordance with accompanying low homology motifs. studies. Furthermore we could show that the understanding Results and conclusions: We developed a method to predict Dedicated gained from an integrated profiling approach outperforms the potential transcriptional regulatory regions and motifs using a Posters analysis of just one level of complexity. combinational approach with support vector machine (SVM) and principal component analysis (PCA). The dataset uses upstream DS2-1-15 sequences divided into 54 gene clusters by Kim’s microarray results. The procedure of our method is as follows: (1) In the Competition between diacylglycerol dependent protein SVM, we examine two methods of encoding for input data, kinase C and insulin receptors for phosphorylation of IRS-1 count encoding and binary encoding with several conditions, e.g. van Riel, Natal1; Wagenmakers, Anton2; Hilbers, Peter1 regions, window size, number of mismatches and direction. To 1Eindhoven University of Technology, Biomedical Engineering, two pulled out of the 54 gene clusters, if the gene clusters are Eindhoven, Netherlands; 2University of Birmingham, School of classified as high accuracy by SVM, the encoding is expected Sport and Exercise Sciences, Birmingham, United Kingdom to contain specific sequences for the transcriptional regulation. (2) The prediction of the regulatory motif is predicted from those Objective: An increase in intramyocellular fatty acyl CoAs candidate encodes using PCA. Our method showed good and diacylglycerol (DAG) is associated with insulin resistance performance to known regulatory regions and predicted new (lipotoxicity). The accumulation of fatty acid metabolism regulatory regions. intermediates can be due to increased delivery from plasma and/ or reduced β-oxidation, caused by mitochondrial dysfunction DS2-1-17 and/or reduced association between mitochondria and lipid droplets. DAG activates several protein kinase C (PKC) Combined theoretical and experimental analysis links isozymes. The activated kinases phosphorylate serine residues isoform-specific contributions of ERK signaling to cell fate on Insulin Receptor Substrate 1 (IRS-1), thereby competing decisions for insulin-induced tyrosine phosphorylation of IRS-1 (tyrP- Schilling, Marcel1; Maiwald, Thomas2; Hengl, Stefan2; Winter, IRS1). To quantitatively test the hypothesis that hyper-serine Dominic3; Kolch, Walter4; Lehmann, Wolf D.3; Timmer, Jens2; phosphorylation of IRS-1 could result in insulin resistance, a Klingmüller, Ursula1 mathematical model of the insulin signaling pathway was used. 1German Cancer Research Center, Systems Biology of Signal Results: DAG dependent serine phosphorylation and insulin- Transduction, Heidelberg, Germany; 2University of Freiburg, induced tyrosine phosphorylation of IRS-1 have been modeled Freiburg Center for Data Analysis and Modeling, Freiburg, as competing mechanisms with equal kinetic parameters. In the Germany; 3German Cancer Research Center, Molecular Structure presence of DAG, the peak and plateau levels of tyrP-IRS1 during Analysis, Heidelberg, Germany; 4University of Glasgow, Beatson insulin stimulation are more than 40% repressed. The maximum Institute for Cancer Research, Glasgow, United Kingdom level of glucose transporter GLUT4 in the cell membrane is 27% reduced. The area under the curve of surface GLUT4 (considered Objective: Cell fate decisions are regulated by the coordinated as a measure of the glucose uptake capacity) is reduced by more activation of signaling pathways such as the mitogen-activated than 30%, which could have a significant impact on glucose protein (MAP) kinase cascade, but contributions of individual tolerance. A different PKC isozyme, PKC-ζ, also competes kinase isoforms are mostly unknown. Furthermore, no quantitative for IRS-1 phosphorylation, but its activity depends on the link between ERK signaling and cell fate decisions has yet been phosphoinositide 3-kinase pathway, downstream of IRS-1. PKC-ζ established. is activated after insulin stimulation, hereby forming a negative Results: By combining quantitative data from erythropoietin feedback. The model was extended to test the hypothesis that (Epo)-induced pathway activation in primary erythroid progenitor the extra PKC-ζ serine phosphorylation would aggravate the cells (CFU-E) with mathematical modeling, we predicted and DAG effect and hence insulin resistance. However, in contrast, experimentally confirmed a distributive ERK phosphorylation the dynamic PKC-ζ feedback significantly reduced the insulin mechanism in CFU-E. Model analysis revealed bow tie-shaped resistance predicted by the model. signal amplification and inherently transient signaling. Sensitivity Conclusions: In conclusion, the model suggests that the analysis predicted that, via a feedback-mediated process, multiple processes competing for IRS-1 phosphorylation hamper increasing one ERK isoform would reduce activation of the a straightforward explanation of insulin resistance based on an other isoform, which was verified by protein overexpression. observed hyper-serine phosphorylation of IRS-1. We calculated ERK activation for biochemically not addressable but physiological ligand concentrations and quantitatively linked these results to cellular proliferation. The results show that

94 ICSB 2008 double-phosphorylated ERK1 controls proliferation in a bell- particular, the role, if any that the Rim pathway plays in alkaline shaped fashion, while activated ERK2 enhances proliferation with adaptation is not well understood. saturation kinetics. Here, mathematical modelling techniques, ODEs, are applied Conclusions: By combining experimental analysis and to these two transcription factors, PacC/Rim101, with the aim mathematical modeling, we show that ERK signaling contributes of gaining theoretical insights into differences in their activation to cell fate decisions in a dose-dependent and isoform-specific mechanisms, and further identifying the role Rim101 plays in manner. With this model we were able to dissect the contributions alkaline adaptation. of activated ERK1 and ERK2, which provided us with a fully Results: The simulation from two models agrees with several predictable quantitative model for cellular proliferation. Thus, we experimental observations. Sensitivity analysis shows PacC established a quantitative link between previously unobservable system is more sensitive to the change of signal strength from signaling dynamics and cell fate decisions. upstream pathway than Rim101 system. We used these models to explore differences in activation mechanisms for the two DS2-1-18 transcription factors. Our predictions suggest that external signal needs to be present for a much longer period, around 10 hours, S. cerevisiae morphology metrics of interference for PacC to be fully activated than for Rim101, which in contrast, metabolites is fully activated within 30 minutes and its level is adjusted at later Castro, Cristiana1; Lopes, Vitor2; Costa Martins, Rui3 time points. 1IBB - Institute for Biotechnology and BioEngineering, University Conclusions: By relating our models to existing biological of Minho, Braga, Portugal; 2ISR - Institute for Systems and knowledge and comparing the downstream gene targets of these Robotics, Instituto Superior Técnico, Lisbon, Portugal; 3Molecular two transcription factors, we speculate that Rim pathway in S. Biology and Environmental Research Center, University of Minho, cerevisiae possibly plays a role in the initial response to metal Braga, Portugal cations and/or nutrient depletion caused by alkaline pH. Whereas in A. nidulans, the PacC pathway seems to be responsible for Objectives: Saccharomyces cerevisiae morphology is known to long term adaptation to alkaline environment. be dependent on the cell physiological state and environmental conditions. On their environment, wild yeasts tend to form DS2-1-20 complex colonies architectures, such as stress response and pseudohyphal filaments morphologies, far away from the ones HGF-mediated PI3K signalling pathway in hepatocytes found inside bioreactors, where the regular cell cycle is observed proliferation: Direct and indirect activation under controlled conditions (e.g. budding and floculating D’Alessandro, Lorenza1; Salazar, Carlos2; Schilling, Marcel1; 3 4 2 colonies). In this research, we interfere with \textit{S. cerevisiae} Maiwald, Thomas ; Hengstler, Jan ; Hoefer, Thomas ; Timmer, Posters from IBB collection: i) wild type; ii) floculant (CA116 strain); and Jens3; Klingmueller, Ursula1 Dedicated iii) non-floculant (YMD90). The colony dynamics was studied by 1German Cancer Research Centre, Systems Biology of Signal morphology metrics when subjected to: i) YP and YPD growth Transduction, Heidelberg, Germany; 2German Cancer Research media (12 hrs at 25oC) innoculated with 1\% (v/v) of ii) aromatic Centre, Heidelberg, Germany; 3University of Freiburg, Freiburg alcohols (ethanol, propanol, isopropanol, phenylethanol); and iii) Center for Data Analysis and Modelling, Freiburg, Germany; amyl radicals (iso-amyl-OH and terthamyl-OH). 4University of Dortmund, Dortmund, Germany Results: After incubation, samples were analysed for population dynamics and morphology statistics by digital microscopy image Objective: By binding to its receptor MET, the Hepatocyte analysis. Morphology was classified as haploid, diploid, meiosis, Growth Factor (HGF) regulates multiple cellular responses, budding and hyphae formation. Data was subjected to relevant including cell proliferation, motility, invasion and protection principal component analysis. PCA (PC1(75.21%), PC2(23.98%)) from apoptosis. In the liver, HGF is produced in response to results show that the wild type is capable of adapting to the most hepatic injury and enhances the entry of hepatocytes into the adverse growth media unde starvation and presence of alcohol, cell cycle. Proliferation of hepatocytes is mainly driven by PI3K isoamyl-OH, terthamyl-OH; and that budding occurs only on YPD signaling. HGF activates PI3K by two different mechanisms: by media. The presence of alcohols tend to decrease reproduction direct binding of PI3K to activated MET and by indirect binding by budding even on YPD (e.g. propanol, isopropanol, ethanol mediated by the adaptor protein Gab1. To dissect the HGF-driven and methanol), being possible to cluster the effect of the different direct and indirect PI3K activation in standardized hepatocyte metabolites on the yeast morphology. culture conditions, we combine the mathematical modeling with Conclusions:Results show that it is possible to intearact with the time-resolved data generated by quantitative immunoblotting. yeast colony morphology by adding communication molecules to Results: The established mathematical model describes the growth media, which trigger morphology changes which are the dynamic properties of the pathway and is divided into not only dependent on starvation. Such opens new opportunities four modules: MET activation represents the input signal; for controlling cell morphology inside bioreactors and PI3K activation is considered by taking into account the two consequently control fermentations, leading to new applications mechanisms of activation; PIP3 formation includes negative and processes in biotechnology. regulators such as PTEN; Akt activation is considered as an indicator of PI3K activity. By parameter estimation we observed DS2-1-19 that the model can reliably describe the data. The theoretical model predictions suggest that PI3K is first activated by direct Comparison of pH adaptation in Aspergillus nidulans and binding, triggering a positive feedback loop by PIP3 production Saccharomyces cerevisiae: A modelling approach that recruits Gab1 to the plasma membrane with the consequent Ke, Ruian1; Stark, Jaroslav2; Haynes, Ken2 amplification of the indirect PI3K activation. Our data indicate 1Imperial College London, CISBIC, London, United Kingdom; that Akt activation is prolonged compared to Met activation, 2Imperial College London, London, United Kingdom supporting the model hypothesis. Conclusions: The results deriving from the combination of the Objective: Fungi grow in a wide range of conditions and generated data and dynamic modeling pathway indicate that adapt to ever-changing external environments by tailoring gene direct and indirect PI3K activations are necessary to activate and expression. An important component of this environment is sustain the signalling pathway. We are currently investigating the external pH. In Aspergillus nidulans members of the Pal signalling dynamics of interactions responsible for the direct and indirect pathway and the transcription factor PacC mediate pH sensing PI3K activation by refining the model taking the protein-protein and transcriptional responses respectively. In Saccharomyces interactions into account. cerevisiae, the homologous genes are members of the Rim pathway and transcription factor Rim101. Though these two pathways are conserved in many aspects, the physiological significance of several notable differences between them, and in

ICSB 2008 95 DS2-1-21 [1] Ventura AC, Sepulchre J-A, Merajver SD (2008) A Hidden Feedback in Signaling Cascades Is Revealed. PLoS Comput Biol Differential regulation of IL-6 signaling pathway in HaCaT 4(3) : e1000041. doi:10.1371/journal.pcbi.1000041 A5 benign tumor keratinocytes and fibroblasts Depner, Sofia1; Scherzinger, Tobias2; Nici, Marco2; Mueller, DS2-1-23 Margareta2 1DKFZ (German Cancer Research Center), Tumor and Dynamic modeling of negative feedback regulation of Microenvironment, Heidelberg, Germany; 2DKFZ (German Cancer TGFbeta-Smad signaling in primary mouse hepatocytes Research Center), Heidelberg, Germany Nickel, Peter J.1; Legewie, Stefan2; Maiwald, Thomas3; Zhu, Qingwei4; Müller, Stephanie1; Bohl, Sebastian1; Frahm, Thomas1; The activated progression promoting tumor microenvironment D’Alessandro, Lorenza1; Meyer, Christoph5; Godoy, Patricio5; is initially induced by a network of tumor derived growth factors/ Bauer, Alexander6; Ueberham, Elke7; Gretz, Norbert5; Gebhardt, cytokines that induce cellular responses in tumor and stromal Rolf7; Hengstler, Jan G.6; Dooley, Steven5; Luo, Kunxin4; Herzel, cells. In a tumor transplantation model of HaCaT skin squamous Hanspeter2; Timmer, Jens3; Klingmüller, Ursula1 cell carcinomas we could demonstrate the functional contribution 1German Cancer Research Center (DKFZ), Systems Biology of of an IL-6 regulated growth factor network to tumor progression. Signal Transduction, Heidelberg, Germany; 2Humboldt University, The network induces tumor cells proliferation and migration as Institute for Theoretical Biology, Berlin, Germany; 3University well as persistent angiogenesis and recruitment and activation of of Freiburg, Freiburg Center for Data Analysis and Modelling, stromal cells. In response to ligand binding the IL-6R activates Freiburg, Germany; 4University of California, Berkeley, United the JAK/STAT signaling pathway in stromal fibroblasts and tumor States; 5University Hospital of Heidelberg, Mannheim, Germany; cells but pathway activation results in the induction of different 6Leibniz Research Center (IfADo), Dortmund, Germany; 7University target genes and triggers different cellular responses in both of Leipzig, Institute for Biochemistry, Leipzig, Germany cell types. This differential target gene response is most likely mediated by a differential kinetics of expression, phosphorylation Objective: In the mammalian liver, transforming growth factor and nuclear localization of STAT proteins (STAT1 and 3) after IL-6 beta (TGFbeta) acts as a mitosis inhibitor and differentiation stimulation in both cell types. Additionally tumor keratinocytes and factor on hepatocytes both in the quiescent state and during Dedicated stromal fibroblasts respond with a different pattern of activation liver regeneration, which occurs after liver injury. Dysregulated Posters for MAP kinases such as Erk1/2. Blockade of one of these IL-6 TGFbeta signaling can lead to liver fibrosis, cirrhosis and induced growth factors (GM-CSF) in tumour keratinocytes led to ultimately cancer. TGFbeta initiates intracellular signal transduction alterations of the IL-6 induced STAT1 and 3 activation kinetics, by activating Smad signaling as well as several other pathways. indicating the existence of an autocrine positive feedback loop in Binding of TGFbeta to its cognate cell surface receptors leads the JAK/STAT signaling pathway between IL-6 and GM-CSF. to the recruitment of latent Smad transcription factors. Active Smad complexes accumulate in the nucleus, where they regulate DS2-1-22 target gene transcription. Various modulators of TGFbeta- Smad signaling have been described. However, their dynamic Prediction of reverse stimulus-responses in signaling interactions and relative contributions to the integrated signaling cascades response in primary hepatocytes remain poorly understood. Sepulchre, Jacques-Alexandre1; Ventura, Alejandra C2; Merajver, Results: In our study, genome-wide expression profiling at the Sofia 2D mRNA level identified potential feedback regulators of TGFbeta- 1University of Nice Sophia Antipolis, Nonlinear Institute of Smad signaling in primary mouse hepatocytes. At the protein Nice, CNRS, Valbonne Sophia Antipolis, France; 2University of level, the stoichiometry of Smad pathway components was Michigan, Department of Internal Medicine, Cancer Center, Ann determined using purified calibrator proteins. To gain insights Arbor, United States into regulatory mechanisms, a dynamic TGFbeta-Smad signaling model was established by employing a systems biology approach Objective: In a systems biology perspective, intracellular combining multiple quantitative immunoblotting data sets with signaling cascades are generally viewed as unidirectional input/ mathematical modeling. After data-based parameter estimation, output devices, characterized by a stimulus-response curve the dynamic model was experimentally validated by testing describing how the state of the activated proteins at the bottom model-derived predictions for previously untested conditions. of the cascade depend on parameters affecting an upstream Conclusion: By dynamic pathway modeling SnoN was identified cycle of the chain. In this paper our goal is to present the novel as the major negative feedback regulator during the early phase concept of reverse stimulus-response that we suggested already of TGFbeta-Smad signaling in primary mouse hepatocytes. in a recent work [1]. We show that a modification of the state SnoN efficiently modulates TGFbeta-Smad-dependent biological of a downstream biochemical cycle in a signaling cascade can responses despite its substoichiometric concentration. bring about a response upstream in the transduction pathway, therefore challenging the widespread idea of unidirectionality in DS2-1-24 signaling cascade. Results: Our finding is first presented in the simplest case Linear motif atlas for phosphorylation-dependent signalling of a bicyclic cascade. In the literature this system is usually Linding, Rune1; Juhl Jensen, Lars2; Yaffe, Mike3; Brunak, Soeren4 represented as a motif comprising 2 cycles and a single arrow 1ICR, Network & Systems Biology Team, London, United linking the activated protein of the first unit onto the second Kingdom; 2EMBL, Heidelberg, Germany; 3MIT, Boston, United cycle. Despite the clear polarity of this system, we show that a States; 4CBS, Lyngby, Denmark variation of the parameters affecting the second cycle can induce an upstream response in variables of the first cycle. This can be Directionality is fundamental to cellular signal progression. This achieved in biochemical relevant conditions. We have named property of protein signalling networks is in part due to linear the obtained relationship as “reverse stimulus-response curve”, motifs that are phosphorylated by kinases and may subsequently and this can be computed numerically, as well as analytically be bound by modular domains such as SH2. To support the estimated. The result is next generalized to a cascade of several proteome-wide characterisation of interactions modulated cycles and to more complex pathways. by post-translational modification, we have developed the Conclusions: The so far overlooked concept of reverse stimulus- NetPhorest resource, an atlas of linear motifs recognised by response increases the potentiality of signaling cascades, as well protein kinases and phosphorylation-dependent binding modules. as the complexity of their dynamics. Our predictions prompt new The resource is maintained by an automated pipeline, which uses experiments in signal transduction networks, and possibly new phylogenetic trees to structure the currently available in vivo and ways to interpret current experimental results. As an interesting in vitro data for training, evaluation and selection of sequence- significance of our study we analyze its implication on signaling based classifiers. Presently the atlas contains gene- or family- pathways with crosstalk. specific classifiers for 179 kinases and 97 phosphotyrosine-

96 ICSB 2008 binding modules. These reveal that tyrosine kinases mutated in sensitive to perturbations for ligand half-life and therefore the cancer have lower recognition specificity than their non-oncogenic formation of signaling-competent ligand-receptor complexes. siblings. We make the atlas available as a community resource Furthermore, we could successfully exploit our model to http://netphorest.info. determine kon of the hyperglycosylated Novel Erythropoiesis Stimulating Protein (NESP) based on quantitative immunoblotting DS2-1-25 data. Therefore, this method presents a tool for the efficient evaluation of Epo derivatives with enhanced bioavailability. Elucidating the mechanisms of initial insulin signalling in Conclusions: Our approach demonstrates the potential of adipocytes using mathematical modelling and complex combining biochemical and mathematical analysis to establish input signals new strategies for the development of refined clinical therapeutics Johansson, Cecilia; Cedersund, Gunnar and to elucidate design principles governing cytokine receptor Insitution of Clinical and Experimental Medicine, Linköping signaling and erythrocyte homeostasis. University, Department of Cell Biology, Linköping, Sweden DS2-1-27 Malfunctioning in the cellular response to insulin, insulin resistance, is a characteristic feature of type 2 diabetes, a Difference and commonality of gene regulatory networks rapidly growing disease which currently affects almost a quarter in ligand-stimulated wild type and drug-resistant MCF-7 of a billion people. The response to insulin is altered already at breast cancer cells the initial signalling steps, and even though the basic skeleton Hatakeyama, Mariko1; Nagashima, Takeshi1; Ikeda, Kazuhiro2; of these interactions is fairly well established, many important Kuroki, Yoko1; Gotoh, Noriko3; Oyama, Masaaki3; Inoue, Satoshi2; features - such as feedbacks, time-scales, and location – remain Kitano, Hiroaki4 uncharacterized. We have employed a systems biology approach 1RIKEN Advanced Science Institute, Yokohama, Japan; 2Research to elucidate more details concerning these uncharacterized Center for Genomic Medicine, Saitama Medical University, features. Our experimental phosphorylation data shows that Saitama, Japan; 3Institute of Medical Science, University of Tokyo, the first activated proteins display an overshoot behaviour, Tokyo, Japan; 4The Systems Biology Institute, Tokyo, Japan where a maximal value is reached within one minute, before an intermediate response is reached within the next 3-5 minutes. Objective: Ligand-stimulated gene expression often works Our mathematical analysis shows that this overshoot observation for fine-tuning and regulation of initial and additional signaling alone is sufficient to rule out the assumption that no feedbacks pathways by positive and negative feedbacks and determines to the receptor, possibly including endocytosis and recycling, are overall long-term cellular kinetics for cell fate control. Therefore, significantly active during the first minutes of the response. To if changes in cellular signaling pathway are constantly acquired Posters further elucidate features of these significantly active feedbacks, among cells, same ligand essentially induces moderately similar Dedicated we perturb the system with more complex input signals, e.g. but different set of genes related to upstream pathway in each cell containing several steps in the insulin concentration. Mathematical subgroup. Based on this strategy, we analyzed effect of estrogen analysis of this data, combined with detailed observations of (E2) and an ErbB receptor ligand, heregulin (HRG) to observe signalling in cellular sub-fractions, gives novel and important regulatory genes that are related to the ER-ErbB cross-talk in mechanistic information, and also allow us to reject further wild type (WT) and tamoxifen (an ER antagonist)-resistant (TamR) assumptions, e.g., that all significantly active feedbacks during MCF-7 breast cancer cells. the first few minutes consists of endocytosis plus recycling. Results: Time-course gene expression analysis was performed Apart from classical cross-validations, an important strength for E2 or HRG-stimulated WT and TamR MCF-7 cells up to 48 and quality-tag of our predictions is that we only consider core hrs (8 time points) using Affymetrix GeneChip. Gene selection predictions, i.e. predictions that are shared for all parameters and followed by principal component analysis of expression data model variations that give an acceptable agreement with the data. indicated a different cause of transcription by E2 and HRG. Pathway enrichment analysis indicated that a differential regulation DS2-1-26 for two ligands involved genes related to “Focal adhesion”, “MAPK signaling pathway”, “Cell communication”, “ErbB signaling Modeling erythropoietin receptor endocytosis identifies pathway” (p<0.05). essential ligand binding properties and rapid recovery of Conclusions: Our experimental analysis indicated that ligand- cell surface receptor stimulation indeed regulates expression of genes related to Becker, Verena1; Schilling, Marcel1; Bachmann, Julie1; Baumann, upstream signaling pathway at high rate. Time-course patterns Ute1; Hengl, Stefan2; Maiwald, Thomas2; Timmer, Jens2; and other analysis indicated that E2 and HRG shares both Klingmueller, Ursula1 common and distinct regulatory machineries for their transcription 1DKFZ, Division Systems Biology of Signal Transduction, that may affect upstream signaling to induce distinct feature of Heidelberg, Germany; 2University of Freiburg, FDM, Freiburg, drug resistance. Such a difference in signaling will be further Germany analyzed by kinetic modeling for investigation of detailed molecular mechanism. Objective: To balance constant self-renewal and rapid adaptation to environmental changes in the hematopoietic system, cytokine DS2-1-28 receptors display several mechanisms to efficiently activate as well as terminate signal transduction. In contrast to receptor Systems biology of vertebrate developmental co- tyrosine kinases, for cytokine receptors such as the erythropoietin expression networks receptor (EpoR) only a minor fraction of receptor protein is Ramialison, Mirana; Henrich, Thorsten; Wittbrodt, Beate; localized to the plasma membrane whereas the majority is Wittbrodt, Joachim retained in intracellular compartments. Therefore, the contribution EMBL, Developmental Biology, Heidelberg, Germany of endocytic downregulation of cell surface receptor for signal attenuation of cytokine receptors remains elusive. Objective: Systems Developmental Biology has recently been Results: Based on quantitative experimental data, we introduced to the scientific community as a computational established non-linear ordinary differential equation (ODE)-based framework to integrate all the data obtained from experimental models describing both ligand-independent and ligand-induced embryology and developmental genomic fields (Bard, 2007). Its endocytosis of the EpoR, the key regulator of erythropoiesis. overall aim is to model developmental processes by deciphering Model predictions were experimentally validated and revealed that the complex interplay between developmental genes that lead signaling through the EpoR is tightly controlled by endocytosis- to structural self-organization. To get insight into this network, mediated ligand consumption, while the population of cell surface we performed a systematic analysis of gene co-expression receptor is rapidly recovered. Sensitivity analysis showed that at the level of a whole vertebrate model organism. We aim the association rate kon of ligand binding is the parameter most to understand the functional and regulatory properties of co-

ICSB 2008 97 expressed genes. other cells of the immune system, thus functioning as antigen- Results: We have collected spatio-temporal expression patterns presenting cells. Several studies have already described the use of medakafish genes by large-scalein situ hybridization on of microarrays to determine the gene expression signature of whole-mount embryos. The data was stored in the Medaka activated DCs upon pathogen detection. However, these studies Expression Pattern Database and annotated using a controlled did not include the complexity related to mRNA translation vocabulary of anatomical terms. From this collection of annotated regulation. Therefore, we aimed to determine the nature of genes, we built a map of co-expressed genes at each specific the mRNAs being translated at various stages of human DC developmental time-point, using terrain maps for data reduction. activation with the help of translational profiling, which is the We discovered that genes sharing similar expression domains sucrose gradient fractionation of polysomal-bound mRNAs (e.g.: proliferative tissues, somites), share common over- combined to microarrays analysis. represented DNA regulatory motifs, which are often conserved Results: As a first step, the rate of protein syntheis in during evolution. Furthermore, we provide evidence that some Lipopolysaccharide (LPS)-stimulated human monocyte- co-expressed groups present a bias in chromosomal position: derived DCs has been quantified during time. An important they are clustered in the same locus, an organisation reminiscent increase in protein synthesis was observed as soon as 1h after of the bacterial operon. LPS-stimulation, peaking at 8h and followed by a continous Conclusions: We have achieved the first systematic discovery decrease between 8h to 24h. Total and polysomal-bound mRNA of synexpression groups (Niehrs and Pollet, 1999), defined as populations were compared in immature (0h) and LPS-stimulated genes sharing similar expression patterns, involved in the same (4h and 16h) DCs with the help of Affymetrix microarrays. A biological process and regulated by the same trans-activating global quantitative analysis of the microarrays signals indicated factors. Our unexpected finding on chromosomal clustering of that 10 to 20% of the genes are translationally repressed at 16h co-expressed genes hints at the existence of novel Global Control post-LPS. No major difference could be observed at 4h post- Regions (Spitz et al., 2003). This work has contributed to the LPS, indicating that translation repression takes place in the “late linking of gene co-expression networks with gene co-regulatory stage” of DC-activation. Biostatistical analysis indicated that networks, through the identification of common regulatory about 300 mRNA molecules are translationally regulated. The elements that are responsible for directing complex spatio- most abundant biological process among the regulated mRNAs temporal gene expression. is protein biosynthesis, indicating the existence of a translational Dedicated feedback loop. Translationally regulated mRNAs involved in Posters DS2-1-30 immunity (MD-2, CD48, CD80, IL-6) have been also identified. Discussion: Our observations highlight the importance of Reconstruction of network models of Snf1 from signaling translation regulation during the immune response. This in yeast information will provide information on the possible absence of Nandy, Subir Kumar; Nielsen, Jens correlation between gene expression and real protein production Chalmers University of Technology, Department of Chemical and in DCs. Our data may also favor the identification of novel DC- Biological Engineering, Gothenburg, Sweden specific gene clusters and protein networks.

Objectives: In eukaryotes the SNF1 protein kinase is conserved DS2-1-32 and implicated in nutritional and environmental stress. One of the best examples is Yeast where SNF1 plays a major central role Systems biology approach reveals novel TNF-alpha in response to glucose starvation. The involvement of protein- signaling crosstalk protein interaction networks including a huge number of protein Hayashi, Kentaro; Helmy, Mohamed; Tomita, Masaru; Tsuchiya, make a complex towards quantitative studies of complex signal Masa; Selvarajoo, Kumar transduction pathways. Institute for Advanced Biosciences, Keio university, Tsuruoka, Results: Here a detailed network reconstruction model of Japan regulatory system will be generated for SNF1 studied in this work. In Yeast, SNF1 is associated with its activity subunit SNF4 Objective: Tumor necrosis factor (TNF) family of cytokines is and also other proteins in complexes. Interaction between these chiefly involved in several important biological processes such two proteins is strongly regulated by the presence of glucose. as tumor necrosis, anorexia, cell proliferation, cell differentiation The reconstructed network models will be represented in both and apoptosis. They are produced by many cell types, e.g. Cytoscape and Cell Designer. These software platforms are not macrophages, monocytes, lymphoid cells and fibroblasts. Among only used for visualization of reconstructed networks but can the 27 members of the TNF superfamily identified, TNF-α is also be used for integration of high throughput experimental data the most studied. Basically, TNF-α binds to the extracellular generated in different studies. TNF Receptor 1 (TNFR1) and triggers the activation of activator Conclusions: All known components of the SNF1 pathways protein-1 (AP-1), nuclear factor-κB (NF-κB) and the induction of were identified together with a large number of new ‘candidate’ many other proinflammatory cytokines such as IL-1 and IL-8, proteins, indicating the successful reconstruction of the SNF1 through the recruitment of adaptor molecules TRADD, TRAF2 and pathways in Yeast. This full pathway can be used to understand TRAF5 to its cytoplasmic domain. Although there are numerous the information of each proteins and reactions involved in SNF1 reports on TNF-α signaling, the dynamic behavior of these by annotation. Hence this integrated network is useful for pathways under intracellular perturbation such as genetic knock- prediction of also other signalling pathways, unknown members in out (KO) still remains poorly understood. the pathway and identification of functional mechanisms from the Results: Here, based on the current topology of the TNF-α different network modules. signaling pathway we developed a dynamical computational model using E-Cell. Utilizing perturbation-response concept, we DS2-1-31 simulated the complex signaling processes from receptor binding until transcription factor (TF) activation of all known signaling Fine-tuning of human dendritic cells regulation revealed by molecules in wildtype and several knock-out (KO) conditions translational profiling (TRAF2 KO, TRAF5 KO and TRAF2/TRAF5 double KO). By Ceppi, Maurizio1; Chaussabel, Damien2; Blankenship, Derek2; comparing the simulation of key MAP kinase c-Jun N-terminal Banchereau, Jacques2; Pierre, Philippe1 kinase (JNK) and NF-κB with time-course experimental data in 1Centre d’Immunologie de Marseille-Luminy (CIML), Marseille, the four conditions, we predict the existence of crosstalk between France; 2Baylor Institute for Immunology Research (BIIR), USA, TRAF5 and IKK complex in TNF-α signaling. Dallas, United States Conclusions: Our systemic approach could help uncover novel features of complex pathway in genetic mutation arising from Objective: Dendritic cells (DCs) are the sentinels of the disease condition. mammalian immune system. The main function of DCs is to process antigen material and present it on the surface to

98 ICSB 2008 DS2-1-33 DS2-1-35

Modeling temporal dynamics and negative feedback Revealing key transcription factors controlled by MAPK regulation of JAK2/STAT5 signaling signaling pathway in the process of pressure overload- Bachmann, Julie1; Schilling, Marcel1; Hengl, Stefan2; Timmer, induced cardiac hypertrophy Jens2; Klingmüller, Ursula1 Hong, Seong-Eui; Kim, Do Han 1German Cancer Research Center, Systems Biology of Signal Gwangju Institute of Science and Technology, Life science, Transduction, Heidelberg, Germany; 2University of Freiburg, Gwangju, Republic of Korea Freiburg Center for Data Analysis and Modeling, Freiburg, Germany Objective: Cardiac hypertrophy (CH), an adaptive response to various mechano•humoral perturbations is accompanied Objective: The development of erythrocytes is tightly regulated by increased protein synthesis, cardiomyocyte size and re- by Erythropoietin (EPO) coordinating proliferation, differentiation organization of sarcomere structure. These changes are linked and survival of erythroid progenitor cells. One of the key pathways to a variety of signaling pathways such as Akt, GPCR, MAPK which is activated after stimulation with EPO, the JAK2/STAT5 and PKC. For the present study, we aimed to investigate the key signaling cascade, is well studied by systems biology approaches, transcription factors (TF) associated with the MAPK signaling however, negative feedback loops regulating duration and pathways activated in pressure overload-induced CH (PO-CH). amplitude of the signal remain poorly understood. Results: To obtain the genes affected by MAPK signaling Results: To investigate how negative feedback regulation pathway in PO-CH, cDNA microarray data from transgenic mice modulates the dynamic behavior of the JAK2/STAT5 pathway, a over-expressing cardiac JNK, p38 and ERK were compared mathematical model was established that captures the effects of with the ones from transverse aorta constriction (TAC) mouse the negative regulating proteins CIS and SOCS-3 as well as the model that induced CH. For the analyses, common differentially tyrosine phosphatase SHP-1. Kinetic parameters were estimated expressed genes (DEG) in the signaling pathways were with quantitative time-resolved data of activated EpoR, JAK2, considered. According to our computational analysis, 88, 47 and STAT5, SOCS-3 and CIS after stimulation with EPO in primary 7 genes in each of the signaling pathways were similarly affected erythroid progenitor cells (CFU-E). The data-based mathematical in the TG and TAC animals. The key transcription factors (TF) model was used to theoretically predict overexpression and were identified by analyzing the upstream TF binding sites in knockdown experiments of SOCS-3 and CIS. Experimental DEGs and non-DEGs. As a result, Klf4, Egr1, NFAT, Oct1, USF1, validation confirmed mathematical simulations. HAND1, SRF, NF1, YY1, FOXO4, RREB1, IRF1 and Myf were Conclusion: Model analysis revealed that SHP-1 is a possible found to over-represent in JNK-mediated CH (P<0.05). On the candidate to control the extent of the initial response whereas other hand, AP-2g, RUNX1, ETS, NFAT, HIF-1, GATA3, ZNF143, Posters SOCS-3 was identified as key negative regulator that modulates C/EBPg, IRF and FoxF2 were found to over-represent in p38- Dedicated the level of long-term phosphorylation dynamics. These studies mediated CH. Interestingly, our data further suggest that the provide a more detailed understanding of the dynamic properties above TFs are under the regulation of other well know TFs (e.g. of the JAK-STAT signaling cascade and could facilitate the c-Jun, ATF2, MEF2, NFAT, STAT3 for JNK; c-Myc:Max, AP2alpha, identification of novel therapeutic targets to abrogate constitutive C/EBPa, CREB, ETS, MEF2, NFAT, NFkB, STAT1 for p38). pathway activation in cancer cells. Conclusions: Our computational study has shown some key TFs associated with MAPK pathway and PO-CH. DS2-1-34 DS2-1-36 Signaling network in Mouse Embryonic Stem Cells Polouliakh, Natalia1; Matsuoka, Yukiko2; Ghosh, S2; Nock, R3; Data analysis and modeling of the mammalian circadian Nielsen, F4; Kitajima, S5; Takagi, A5; Aisaki, K.I5; Kanno, J5; Kitano, clock Hiroaki6 Westermark, Pål1; Kramer, Achim2; Herzel, Hanspeter1 1Sony Computer Science Laboratories Inc., Tokyo, Japan; 1Institute for Theoretical Biology, Berlin, Germany; 2Charité 2Systems Biology Institute, Tokyo, Japan; 3CEREGMIA – UAG, Universitätsmedizin, Berlin, Germany Tokyo, Japan; 4École Polytechnique, Tokyo, Japan; 5National Institute of Health Sciences, Tokyo, Japan; 6Systems Biology Objective: Negative feedback loops regulating clock genes Institute, Sony Computer Science Laboratories Inc., Tokyo, Japan constitute the molecular basis of endogenous circadian oscillations. About 20,000 synchronized neurons in the Successful reprogramming of differentiated somatic cells into suprachiasmatic nucleus (SCN) control daily rhythms of a pluripotent state in human and mouse has become a major physiology, metabolism, and behavior. We have previously breakthrough for the potential applications in regenerative combined mathematical modeling with bioluminescence medicine. Such induced pluripotent stem cells (iPS) were reported recordings of fibroblasts to uncover molecular control systems to be generated from dermal fibroblast and other somatic cells by governing oscillation period [1], and differences in chronotype, retroviral transduction of transcription factors, such as: Oct3/4, i.e. being a morning or a night person [2]. Recently, we Sox2, Klf4, c-Myc (Yamanaka, 2007). Auxiliary role of two other predicted that single SCN neurons are damped rather than self- factors, namely, LIN28 and Nanog was also referred in different sustained oscillators [3] based on properties of synchronized literature. Although the comparison between iPS cell and ES cell cell populations. Here, we present a data analysis of single-cell has shown a fraction of genes differently expressed (≈1,500), bioluminescence recordings combined with stochastic oscillator there is a big possibility of sharing a common gene network models, revealing fundamental properties of the fibroblast and involved in ES-iPS self-renewal and maintenance of pluripotency. SCN neuron oscillators and bridging the gap between single-cell In this work we aim to produce a signaling network map of behavior and population properties. embryonic mouse cell (Embryo and Fetus) based on microarray Results: We considered two possibilities for the single-cell expression data and transcription regulation network prediction. circadian clock, a damped oscillator driven by noise or a self- For clustering analysis, an ‘in house’ Gene Geometric spectral sustained noisy oscillator, and fit theoretical autocorrelations to Clustering Tool (AG|GCT) is used and motif discovery is done by those of bioluminescence time-series. We extracted damping, scanning target dataset with known motifs matrices and ab initio noise level and amplitude and used these to classify different prediction techniques. Particularly a hidden markov model (HMM) mutants. Even for the self-sustained case, fits hinted at a for 9 known ‘master’ transcription factors was constructed basing unimodal probability distribution and thus suggest an essentially on another human experimental dataset of 453 genes (Boyer damped oscillator. We calculated resonance curves and et. al, 2005) and applied to the mouse dataset under study. The entrainment phases given the fitted models. Comparing these network will be finalized with CellDesigner software package and with single-cell bioluminescence recordings from synchronized presented for the discussion on the conference. neurons in intact SCN slices favored the purely damped model. Conclusions: Biochemical oscillators driven by biochemical

ICSB 2008 99 noise have long been predicted, and we here present an analysis DS2-1-39 of circadian clock bioluminescence time-series supportive of this notion. It was possible to extract fundamental properties from Different roles of redundant β-subunits of Snf1 protein relatively short, noisy time-series from which synchronization kinase in Saccharomyces cerevisiae properties could be deduced and mutants could be classified. Zhang, Jie [1] K. Vanselow et al. (2006) Genes Dev 20:2660-2672. Chalmers University of Technology, Gothenburg, Sweden [2] S. Brown et al. (2008) PNAS, 105:1602-7. [3] J.C. Locke et al. (2008) BMC Syst Biol, 2:22. Objective: The conserved Snf1/AMPK (AMP-activated protein Kinase) family is one of the central components in nutrition DS2-1-37 sensing and regulation of metabolism in eukaryotes. In mammals, AMPK is a potential drug target for metabolic disorders such as A mathematical model of the regulation of AMPK activation obesity and type 2 diabetes. In yeast S. cerevisiae, Snf1 plays Bittig, Arne T.; Wolkenhauer, Olaf important roles in energy homeostasis, as well as several other University of Rostock, Systems Biology and Bioinformatics Group, processes such as stress resistance, invasive growth and ageing. Rostock, Germany Snf1 is composed of a catalytic subunit Snf1p, one of the three scaffolding β-subunits (Gal83p, Sip1p or Sip2p) and a regulatory Objective: AMP-activated protein kinase (AMPK) regulates many subunit Snf4p. The β-subunit was reported to mediate subcellular pathways related to the cellular energy metabolism. How AMPK localization of Snf1 under derepressing conditions and the itself is regulated is not quite clear yet, but it seems to differ from objective is to study their roles in signal transduction. the regulation of its yeast analogue SNF1. Here, we focus on Results: We constructed single, double and triple beta-subunit(s) the activation mechanism recently proposed by Xiao et al. for deletion mutants and characterized their physiology using batch mammalian AMPK. We devise a corresponding mathematical fermentation. Microarray experiments were performed on single model and investigate the properties of its parameters even deletion mutants (Δsip1, Δsip2 and Δgal83) to investigate the before experimental data to fit the model to are available. cellular transcriptional response to the loss of each specific Results: We perform mathematical optimisation of the beta-subunit. Batch experiments showed that deletion of any parameters with respect to the overall system behaviour and beta-subunit(s), i.e., single, double or triple, didn’t cause any Dedicated analyse the resulting models using sensitivity analysis, principal observable growth defects on glucose, but prolonged diauxic shift Posters component analysis, and other systematic tests. We find that to different extent, suggesting that the expression of enzymes in fewer degrees of freedom than in the original model are needed corresponding metabolic pathways decreased due to the loss to achieve the observed and hypothesised system behaviour, of Snf1 kinase activity. Δsip1 and Δsip2 had similar transcription and that inhibition of AMPK*-dephosphorylation by AMP is no profiles and deletion of Gal83p caused bigger changes than necessary condition for it. deletion of Sip1p or Sip2p, in terms of both the number of genes Conclusion: While actual experimental data remain indispensable affected and the fold-change of expression level. for a definitive model, mathematical tools for system and data Conclusions: β-subunit of the Snf1 protein kinase in yeast S. analysis reveal interesting trends and properties of the system. cerevisiae is redundant. However, there are still some differences, for example, Gal83p is more important than the other two DS2-1-38 in carbon metabolism, which might be due to their distinct subcellular localization and therefore different signal transduction The role of dynamic stimulation pattern in the analysis of mechanisms. bistable intracellular networks Millat, Thomas1; Srenath, Sree N.2; Soebiyanto, Radina P.2; Avva, DS2-1-40 Jayant2; Cho, Kwang-Hyun3; Wolkenhauer, Olaf1 1University of Rostock, Institute of Computer Science, Rostock, Unraveling the cell-cell communication in the pancreas Germany; 2Case Western Reserve University, Complex Systems tumor microenvironment Biology Center, Cleveland, United States; 3Korea Advanced Rogon, M.Z.1; Szabowski, A.2; Giese, N.3; Busch, H.1; Eils, R.1 Institute of Science and Technology (KAIST), Department of Bio 1German Cancer Research Center, Div. of Bioinformatics, and Brain Engineering and KI for, Daejeon, Republic of Korea Heidelberg, Germany; 2German Cancer Research Center, Div. of Signal Control and Tumor Growth, Heidelberg, Germany; Bistable systems play an important role in the functioning of 3University of Heidelberg, Hospital, Heidelberg, Germany living cells. Depending on the strength of the necessary positive feedback one can distinguish between (irreversible) “one-way Objective: Pancreatic cancer is the leading cause of cancer switch” or (reversible) “toggle-switch” type behaviour. Besides death, patients with carcinoma of the exocrine pancreas have the well-established steady-state properties, some important especially dismal prognosis with a five year survival rate of characteristics of bistable systems arise from an analysis of their <1% and a median survival of 4-6 months. The only successful dynamics. We demonstrate that a supercritical stimulus amplitude therapies include surgical intervention or chemotherapy. One is not sufficient to move the system from the lower (off-state) to of the defining features of pancreatic cancer is the unique the higher branch (on-state) for either a step or a pulse input. A microenvironment consisting of pancreatic stellate and switching surface is identified for the system as a function of the inflammatory cells, macrophages, nerve fibers and marrow ] initial condition, input pulse amplitude and duration (a supercritical derived stem cells, which nourish cancer cells, facilitate their signal). Towards this end, we investigate and characterize the metastatic potential and convey radio- and chemoresistance. This role of the duration of the stimulus. Furthermore we show, microenvironment is established by paracrine cell communication that a minimal signal power is also necessary to change the and defined via the cytokine patterns taken up and secreted steady state of the bistable system. This limiting signal power is by each cell. Our aim is to identify key communication players independent of the applied stimulus and is determined only by by an in silico reconstruction and dynamic simulation of the systems parameters. These results are relevant for the design of gene regulatory networks defining the cytokine-pattern related experiments, where it is often difficult to create a defined pattern information processing of each cell type. for the stimulus. Furthermore, intracellular processes, like receptor Results: We have recently shown that the long-term cellular internalization, do manipulate the level of stimulus such that behavior is captured and reflected in the cell’s gene expression level and duration of the stimulus is conducive to characteristic kinetics. Therefore, we set up a high throughput data analysis and behavior. modeling pipeline to describe the establishment and dynamics of pancreas tumor microenvironment in silico. Large scale, time- resolved expression profiles are used to identify significal gene kinetics. We perform a gene ranking, filtering and clustering based on the gene kinetic profiles and finally perform modeling of the gene regulatory network using a continuous time recurrent neural

100 ICSB 2008 network approach. key cytokines from intracellular phospho-protein measurements. Conclusions: Our global gene expression analysis, ranking and Cytokine production was correlated with phospho-proteins modeling enabled us to capture and predict network dynamics in several pathways, suggesting a high degree of crosstalk in and to define the dynamic interactions between genes of interest inflammatory signaling. to establish practically useful drug targets. Conclusion: The challenges of handling such large data sets motivated the development of the DataRail toolbox [http://code. DS2-1-41 google.com/p/sbpipeline/wiki/DataRail]. DataRail is an open source MATLAB toolbox for managing, transforming, visualizing, Dynamic modeling of time-course microarray data by and modeling data, in particular the high-throughput data estimating hidden transcriptional activity profiles encountered in Systems Biology. The analysis of the U937 data Tiemann, Christian1; Moon, Simon2; Stark, Jaroslav2; Starmans, was facilitated by DataRail’s data-driven modeling tools. Maud3; Lambin, Philippe3; Hilbers, Peter1; van Riel, Natal1 1Eindhoven University of Technology, Biomedical Engineering, DS2-1-43 Eindhoven, Netherlands; 2Imperial College London, Mathematics, London, United Kingdom; 3Maastricht Radiation Oncology Identification of substrate proteome for the Akt1 kinase (MAASTRO), Maastricht, Netherlands Yang, Chin-Rang University of Texas Southwestern Medical Center at Dallas, Objective: It is a challenge to reveal and understand gene Simmons Comprehensive Cancer Center, Dallas, United States networks. Microarrays are a great utility for this, as with microarrays it is possible to monitor gene expression of Objective: AKT is a hub kinase of signaling cascades in thousands of genes simultaneously. This presents new challenges determining cellular fate in response to stimulation, and driving for the development of methods for the analysis of tremendous cells toward cancer-prone phenotype, i.e. anti-apoptotic and pro- amounts of data. Last decade, many different methods have been proliferative processes. To further explore the AKT downstream proposed. However, often they are limited and inaccurate. The pathways, we proposed a more comprehenisive protein reason for this is that these methods neglect important biological microarray study of AKT substrate proteome. Akt1, one isoform of factors, like transcript degradation rates, sensitivity of genes to the AKT family, is a serine-threonine protein kinase downstream transcription factors and the activity of transcription factors itself. of the phosphoinositide 3-kinase (PI3K) pathway. So far, more In this work, a method has been developed, using a different than seventy proteins have been reported to be phosphorylated approach. A dynamic ordinary differential equation model has by Akt1. been used, which takes account for biological factors. Model Results: In this project we exausted the detection of Akt1 parameters are estimated using time-course microarray data. substrates using the human protein microarrays containing Posters Results: The developed method takes the data of an arbitrary 8000 recombinant human proteins obtained from Invitrogen’s Dedicated number of microarray experiments as input and estimates UltimateTM ORF (open reading frame) collection. Arrays were the underlying transcriptional activity profile of each gene and incubated with purified Akt1 kinase and 33P-ATP in duplicate. All produces a ranked list of genes that behave most consistent possible kinase substrates were evaluated by their Z-Factor rank in the different experiments. The method is based on a within the array and compared to the negative control. Forty-eight mathematical model of gene transcription. First, to test the putative Akt1 kinase substrates were identified as statistically method, artificial data has been generated, based on a known significant and all were confirmed through visual inspection of activity profile. Subsequently, to further test the method data of array images. Five known Akt1 substrates (CHUK, WEE1, ACLY, three microarray time course experiments, in which hypoxia has PFKFB2, and BRAF) were identified. Receptor tyrosine kinases been induced in different cancer cell lines, has been used. (RTKs) and membrane receptors are enriched. Several SH2/SH3 Conclusions: The method is able to reproduce artificial activity domain-containing tyrosine kinases on the list are also known to profiles correctly. Furthermore, the estimated activity profiles of interact with transmembrane receptors and/or PI3K. Very little is the hypoxia dataset look biologically convincing. The produced known about membrane-bounding proteins are substrates for ranked list of genes is promising as a lot of the top genes have Akt1. These findings pose new questions about feedback loops been related to cancer or hypoxia in other studies. With this and/or signal termination, functions typically ascribed to protein method it is possible to include biological factors and it is easy phosphatases. adaptable to specific cases. Furthermore, it is easy to include Conclusions: In summary, the results of these assays point to newly obtained microarray data. a wide range of functional roles for Akt1, including growth factor receptor signaling, cell cycle and proliferation, cytoskeleton, DS2-1-42 sumoylation, neuronal growth and development. Our results enrich the repertoire of the knowledge regarding the function of Using the datarail toolbox to analyze inflammatory signals AKT in the complex cellular network and the significantly identified in U937 macrophages candidates are further validating both in vitro and in vivo. Goldsipe, Arthur1; Saez-Rodriguez, Julio2; Espelin, Christopher3; Sorger, Peter2; Lauffenburger, Douglas1 DS2-1-45 1MIT, Department of Biological Engineering, Cambridge, MA, United States; 2Harvard Medical School, Department of Systems The interplay between caveolae/raft-mediated endocytosis Biology, Boston, MA, United States; 3Pfizer Research Technology and focal adhesion signaling Center, Cambridge, MA, United States Stergiou, Lilli; Pelkmans, Lucas ETH Zurich, Molecular Systems Biology, Zurich, Switzerland Objective: In order to better understand the signaling events surrounding inflammation, high-throughput techniques were The process of endocytosis plays a crucial role in maintenance of used to measure the dynamics of intracellular phospho-protein cell physiology and homeostasis, by regulating a broad spectrum levels and cytokine release from U937 cells treated with a various of signaling processes, such as cell migration, cell adhesion, inflammatory stimuli. The U937 cell line serves as a model for growth, proliferation and cell polarity. Viruses and ligands for monocytes and macrophages, both of which are present in receptors have been proven to be a useful tool to dissect inflamed tissue. The cells were treated with lipopolysaccharide membrane trafficking routes and follow their intracellular journey (LPS), pro-inflammatory cytokines, and small-molecule inhibitors, to destination compartments, therefore measure endocytic activity resulting in a data set of more than 100,000 measurements. of cells. Results: Principal component analysis (PCA) revealed that most We used siRNA screens in large populations of human cells to of the measured phospho-proteins and cytokines contained only understand how caveolae/raft-mediated endocytosis regulates one dynamic component, although a few species had distinct focal cell adhesion signaling and how these signaling events, early and late components. Partial least-squares regression in turn, regulate membrane trafficking events via caveolae. We (PLSR) was subsequently used to develop predictive models of utilized Simian Virus 40 (antibody traceable) and fluorescently

ICSB 2008 101 labeled cholera toxin B (ChTxB) subunit as a functional readout of the glucose sensing and signaling pathway. This stimulation of our infection experiment or endocytic assay, respectively. of the pathway activates the associated regulatory network. Additional markers were used to monitor the dynamics of According to the direction of the change in the glucose level, caveolae and the status of integrin-based focal adhesions. With genes involved in the utilization of glucose and other carbon image analysis software we extracted features from images of sources are activated or repressed. In yeast S. cerevisiae, some siRNA-treated cell populations acquired by low-resolution light of the proteins involved in the glucose repression pathway microscopy. Support vector machines were used to classify single are clearly identified. A key element of the glucose repression cells into phenotypic groups. pathway, Snf1 protein kinase is composed of a catalytic subunit With this information we generated phenotypic networks of the Snf1p, a regulatory subunit Snf4p and a scaffolding β-subunit genes identified to influence caveolae/raft-mediated endocytosis (Sip1p, Sip2p or Gal83p). Snf1 protein kinase also plays an or the stability of focal adhesions. We observed that internalization important role in also several other processes such as stress of SV40 and ChTxB is regulated by a distinct but overlapping set resistance, invasive growth and ageing. Due to its key role, it is of integrin family members. Moreover, the intracellular trafficking important to predict how signaling mechanism is affected in the of SV40 and ChTxB is under the control of a set of Rab and absence of this complex. Rho GTPase family members, some of which are common Results: In this study, the genome wide expression data from regulators of both routes. Interestingly, a separate group of snf1Δ, snf4Δ and snf1Δsnf4Δ mutants are mapped onto pre- GTPases is required for a normal focal adhesion pattern. Our processed protein-protein interaction network of yeast. All the results corroborate the need of integrity of focal adhesions for linear paths starting from the glucose importer protein Hxk2p and a functional caveolae/raft endocytic pathway, and point to a ending at the transcriptional repressor of gluconeogenic genes divergence of this pathway to potentially accommodate entry and Mig1p were identified via NetSearch algorithm. The linear paths transport of various cellular components. were then scored based on the expression levels of the genes encoding the proteins involved. Thus, the key signaling elements DS2-1-46 of the significantly responsive linear paths were determined. On the other hand, the transcriptome data from these mutants were Protein design as a tool for dissecting signal transduction mapped onto TF regulatory network of yeast. pathways Conclusion: A complete view of the glucose repression Dedicated Kiel, Christina mechanism in the absence of the two constituent proteins of the Posters CRG, Systems Biology, Barcelona, Spain Snf1 protein complex was predicted and putative elements with similar roles to this complex were proposed. Prediction of protein-protein interactions and protein design are important tools in systems biology. In silico predictions DS2-1-48 of proteinprotein interactions can complement large scale experimental studies on determining protein-protein interactions. The variable regulatory mechanism of the calcineurin/NFAT Protein design allows for rationally changing biophysical signaling pathway depending on stimulation strength properties (affinities, or association and dissociation rate Shin, Sung-Young1; Lee, Keun Woo2; Choo, Sang-Mok3; Kwon, constants) of protein complexes, followed by the analysis of the Ki-Sun2; Cho, Kwang-Hyun1 consequence for the system, e.g. a particular pathway. Designing 1Korea Advanced Institute of Science and Technology (KAIST), protein variants which specifically bind to one partner protein Department of Bio and Brain Engineering, Deajeon, Republic opens the possibility to investigate the functional importance of Korea; 2Korea Research Institute of Bioscience and of a specific protein-protein or ligandreceptor interaction Using Biotechnology, Laboratory of Cell Signaling, Deajeon, Republic structural information and protein design tools we have done of Korea; 3University of Ulsan, School of Electrical Engineering, homology modelling and energy calculations to predict the Ulsan, Republic of Korea interaction between 20 Ras subfamily proteins with 50 putative Ras binding domains. To validate this network we have cloned Objective: Calcineurin/nuclear factor of activated T cell (CaN/ six not so well characterized Ras binding domains (RBD) and NFAT) signaling pathway plays crucial roles in the development two Ras proteins. These, together with previously described of cardiac hypertrophy in response to pathological stimuli. RBD domains, Ras and Rap proteins have been analyzed in 70 However, the regulatory mechanisms of this pathway are not fully pull-down experiments. Comparing our interaction network with understood; for example, the contradictory roles of MCIP1. The these and previous pull-down experiments (total of 150 cases) aim of our study is to investigate the functional role of MCIP1 on shows very high accuracy for distinguishing between binders a cellular level and analyzed the regulatory mechanism of CaN/ and non-binders (~0.80). One important question in signal NFAT signaling for varying stress stimuli through combined efforts transduction and systems biology is, whether the magnitude of of cell-based experiments, mathematical modeling, and computer signal transduction is dependent on the complex affinity alone, or simulations. whether individual rate constants of association and dissociation Results: We have revealed that MCIP1 expression decreases the important are important as well. For this aim protein complexes NFAT activity in a dose-dependent manner while RNA interference have been designed which have similar affinity, but different rate of MCIP1 increases the NFAT activity regardless of isoproterenol constants of association and dissociation. These complexes are stimuli in C2C12 myoblasts, which suggests that MCIP1 functions further analyzed in a cellular assay to monitor the effects on signal only as a calcineurin inhibitor at a cellular level. We observed that transduction in vivo. The rate constant of association could be extracellular signal-regulated kinase 5 (ERK5) elevated the NFAT important in systems which get activated by short pulses, and activity in a dose-dependent manner with additional increase here signal transduction might depend on the speed of effector under isoproterenol stimulation. This, consistent with previous proteins binding to their downstream targets (e.g. in the EGF- reports, implies that ERK5 facilitates the CaN/NFAT signaling MAPKkinase pathway). through phosphorylation of MCIP1 under stress stimuli. We also observed that the NFAT activities were gradually increased DS2-1-47 by isoproterenol treatment up to 50 µM, but decreased above this limit. To explain this biphasic response, we propose a new Integration of interactome and transcriptome data to reveal hypothesis that atrogin1 and MuRF1, known ubiquitin ligases for glucose repression pathway in S. cerevisiae calcineurin, are induced under strong stress stimuli and decrease Hasdemir, Dicle1; Onsan, Z. Ilsen1; Nielsen, Jens2; Kirdar, Betul1 the protein level of calcineurin, leading to the suppression of 1Bogazici University, Department of Chemical Engineering, the NFAT activity. This hypothesis has been verified through in Istanbul, Turkey; 2Chalmers University of Technology, Gothenburg, silico simulations and also supported in part by the previous Sweden experimental results. Conclusions: We have revealed that the NFAT activity is primarily Objective: Due to glucose’s role as a primary signaling molecule, modulated by ERK5 (facilitative) and MCIP1 (inhibitory) under mild a change in glucose level immediately stimulates certain elements stimuli, while it is mainly modulated by atrogin1 and/or MuRF1

102 ICSB 2008 (inhibitory) under strong stimuli. This suggests that the primary into several groups by using literature based knowledge and a regulatory proteins of the CaN/NFAT signaling pathway vary statistical time-series model, e.g, a state space model. For the according to the levels of stimuli. predicted factors in the EGF signaling network, we are on the way of confirmation at protein levels in order to reflect them in the DS2-1-49 simulation model. Moreover, the time course analysis revealed novel candidates of biomarkers for prediction of the efficacy of the The influence of the 3’ untranslated region (3’UTR) on drug or for prognosis of lung cancer. quantification of C-terminally tagged proteins in yeast Conclusions:Our next step is to integrate the time-course data Norbeck, Joakim1; Lind, Kristina2 and the simulation model using a new statistical methodology 1Chalmers Tekniska Högskola, Chemical and Biological called data assimilation for extracting novel regulations in the Engineering, Göteborg, Sweden; 2TATAA Biocenter, Göteborg, EGF signaling network. Our approach would certainly advance Sweden personalized medicine in the near future.

Objective: Post-transcriptional regulation is emerging as an DS2-1-51 important player in determining gene/protein expression. Many of the regulatory processes act via the 3’ untranslated region Single-cell dynamics of TRAIL-sensitivity in cancer cells (3’UTR) of the mRNA. The large scale TAP- and GFP-tag Flusberg, Deborah; Spencer, Sabrina; Albeck, John; Millard, collections in yeast all use the same 3’UTR, thereby introducing Bjorn; Sorger, Peter a potential source of error in protein quantification. We have Harvard Medical School, Department of Systems Biology, Boston, evaluated the effect of varying the 3’UTR. United States Results: We have developed an immuno qPCR based approach for accurate and sensitive quantification of the TAP-tag. The TAP- Objective: TRAIL (TNF-Related Apoptosis Inducing Ligand) is tag was then used as a reporter gene for studying the effect on currently in clinical trials due to its selectivity in targeting cancer gene/protein expression of varying the 3’UTR. Yeast strains with cells for apoptosis with little effect on normal tissue. However, reporter plasmids in which the TAP-tag expression was governed cancers vary widely in their response and often exhibit TRAIL- by the choice of the 3’UTR were grown in three different media resistance. The goal of this work is to use a combination of and expression of TAP-protein and mRNA was quantified. It was computational modeling and single-cell assays to examine clear that each 3’UTR mediates a unique expression profile, with the heterogeneous nature of TRAIL-sensitivity among cancer effects on mRNA levels and/or translation efficiency. We also populations, as well as within a population of individual cancer found evidence for a clear 3’UTR specificity of mutants involved in cells. deadenylation (pop2Δ) and decapping (lsm1Δ). Finally, we present Results: Differences in TRAIL-sensitivity are reflected at Posters evidence that question the notion that Puf-proteins will always act the single-cell level, where a percentage of a cell population Dedicated in conjunction with Pop2. succumbs to apoptosis and the remainder survives. Sensitizing Conclusions: We have developed a system suitable for highly agents shift this balance in favor of apoptosis; however, variability quantitative studies of 3’UTR dependent gene expression. Initial still exists in the timing of death of individual cells. Furthermore, results indicate that many conclusions on gene expression from an analysis of clonal cell populations indicates a significant the large scale TAP- and GFP-collections will be incorrect due to nongenetic component to this variability. A differential equation- the omission of 3’UTR dependent post-transcriptional regulation. based model built in our lab predicts the switch-like behavior of cells undergoing apoptosis in response to TRAIL, and shows that DS2-1-50 the variable waiting period leading up to cell death occurs largely prior to MOMP (mitochondrial outer membrane permeabilization). Identification of new biomarkers and molecular targets of In HeLa cells, this pre-MOMP phase is characterized by rising lung cancers by systems biology approach caspase-8 activity, followed by MOMP and subsequent cell death Yamauchi, Mai1; Yamaguchi, Rui2; Nagasaki, Masao2; Shimamura, (measured by FRET reporter and flow cytometry assays). Breast Teppei2; Imoto, Seiya2; Saito, Ayumu2; Ueno, Kazuko2; Hatanaka, cancer cells exhibit heterogeneous dynamics of TRAIL-induced Yousuke2; Yoshida, Ryo3; Okeguchi, Kazuyuki3; Kohno, Takashi4; caspase-8 activity prior to MOMP, with resistant cells undergoing Yokota, Jun4; Miyano, Satoru2; Gotoh, Noriko1 an indefinite pre-MOMP waiting period. Sensitizing agents such 1Institute of Medical Science, University of Tokyo, Division of as paclitaxel and EGFR inhibitors shorten the pre-MOMP phase, Systems Biomedical Technology, Tokyo, Japan; 2Institute of leading to earlier cell death. We have begun to investigate Medical Science, University of Tokyo, Human Genome Center, the sources of pre-MOMP variability in different cell lines by Tokyo, Japan; 3Institute of Statistical Mathematics, Tokyo, Japan; comparing pre-MOMP caspase-8 activity dynamics. 4National Cancer Center Research Institute, Biology Division, Conclusions: We find that cell-to-cell variability plays a role in Tokyo, Japan a cancer’s susceptibility to apoptosis. The signaling dynamics upstream of MOMP contribute to this variability, leading to Objective:The epidermal growth factor receptor (EGFR) family differences in apoptotic sensitivity at both the single-cell and of receptor tyrosine kinases lies at the head of complex signal population levels. transduction cascades for cell proliferation and migration. Aberrant activity of members of this receptor family plays a DS2-1-52 key role in tumorigenesis including non small cell lung cancer (NSCLC). Although small-molecule tyrosine kinase inhibitors A dynamic functional network of human regulatory T-cell (TKIs) that target the kinase domain of EGFR were developed, genes there are still problems to distinguish between drug-sensitive and He, Feng drug-insensitive lung cancers in patients. It is assumed that this is Helmholtz Center for Infection Research, Braunschweig, Germany partly ascribed to different levels of oncogenic addiction by EGFR family tyrosine kinases in cancer tissues, for which the molecular Objective: Human FOXP3+CD25+CD4+ regulatory T (Treg) cells mechanisms are still unclear. In order to clarify the mechanisms of play a dominant role in the maintenance of immune homeostasis. addiction to EGFR signaling, we studied dependency of normal However, a systematic analysis revealing genes involved in the cells to EGF signaling network at systems levels. suppressor function of human Treg cells and providing the genetic Results:Using small airway epithelial cell (SAECs), we architecture of the functional network is still lacking. constructed simulation model of EGF signaling by using Cell Results: We have measured the genome-wide expression of Illustrator. To build the model, we analyzed time-course events of 38,500 genes by performing a time-series analysis at ~20 time transcription that are initiated by the EGF with or without gefitinib, points during the process of Treg cell activation. These time-series EGFR-targeted TKI. Newly obtained long-term time-course gene data are then used as part of a reverse-engineering strategy to expression profiles after these stimulations (19 time points in 48 identify the core dynamic functional network of human Treg cells. hours) reveal enormous transcript actions that can be divided Many key candidate genes are directly identified by analyzing the

ICSB 2008 103 hubs of the functional network. Strikingly, among these genes, DS2-1-54 CTLA4, STAT3 and TNFα, already known to play a role in Treg suppressor function, are ranked at top positions. Moreover, a Quantitative morphodynamic analysis of FRET time-lapse combination of the Treg functional network with results obtained imaging from overexpressing GARP, a recently discovered key gene in the Tsukada, Yuki1; Aoki, Kazuhiro2; Nakamura, Takeshi2; Sakumura, human Treg suppressor function, discloses a GARP-dependent Yuichi1; Ishii, Shin1 functional network. In this GARP-dependent network, several 1Graduate School of Information Science, Nara Institute of genes reported to be important for Treg suppressor function Science and Technology, Nara, Japan; 2Graduate School of are identified. Three out of 11 genes with a functional linkage to Medicine, Kyoto University, Department of Pathology and Biology GARP are known to be involved in human autoimmune diseases of Diseases, Kyoto, Japan or Treg suppressor function. Using time-series real-time PCR measurements a significant expression difference of several novel Objective: Dynamic property of regulation of morphological candidate genes is found between Treg and T effector cells. In changes by molecular signaling is being unclear. This study aims addition we determined the relative order of the onset of activation at exploring the spatio-temporal relationship between activity of the different signaling pathways following Treg cell stimulation. of Rho family small GTPases and morphological change in a Unexpectedly this revealed the activation of the circadian rhythm quantitative manner. pathway before peaking of FOXP3-expression. Results: We developed an algorithm called edge evolution Conclusion: The data reported represent the first high-quality tracking (EET) for quantifying cell boundary movements including dynamic functional gene-gene network of human Treg cells. arbitrary complex boundary transitions. Time-lapse fluorescence resonance energy transfer (FRET) images were analyzed by EET. DS2-1-53 Conclusions: EET enables us to trace local edge extension and contraction by defining subdivided edges and their Genetic and gene expression interactions for inference of correspondence in successive frames. Then, it allows biological pathways in yeast Saccharomyces cerevisiae investigation for cross-correlation between local morphological Mattiazzi, Mojca1; Curk, Tomaz2; Kaferle, Petra1; Zupan, Blaz2; change and local intensity of fluorescent signals considering Petrovic, Uros1 time shifts. By applying EET to FRET images of the Rho-family Dedicated 1Jozef Stefan Institute, Department of Molecular and Biomedical GTPases (Rac1, Cdc42 and RhoA), we examined the cross- Posters Sciences, Ljubljana, Slovenia; 2University of Ljubljana, Faculty of correlation between the local area difference and the activity Computer and Information Sciences, Ljubljana, Slovenia of the GTPases. The correlation changed with time shift as expected, so that the correlation peak appeared with 6-8 minutes Objective: Inference of biological pathways is the main aim of the time shift of morphological changes precede to the Rac1 or exploratory genome-wide analyses. Generation of potentially new Cdc42 activities. Our method enables to statistically investigate and useful hypotheses requires different types of experimental the relationship between cell morphological changes and its data, as well as new computational approaches to combine these regulators considering time shifts of the relationship, thus, it data. Gene expression and genetic interaction data are two of the increases the value of time-lapse imaging data to understand most informative types of data in functional genomics. dynamics of various cellular functions. Results: We here propose a rationale and a simple computational method which, rather than correlating findings from these two DS2-1-55 data types, uses each of them independently to identify the components (regulators, effectors) of molecular pathways, and Imaging and mathematical analysis of heterogeneous and then integrates the results into a single network. We show an dynamic gene transcription in living pituitary cells example of this approach for reproducing the model for the action Harper, CV1; Finkenstadt, B2; Rand, DA3; Friedrichsen, S4; of the drug rapamycin on its molecular targets and pathways in Semprini, S5; Mullins, JJ5; Davis, J6; White, MWH1 yeast cells. While robust techniques to accurately determine gene 1University of Liverpool, School of Biological Sciences, Liverpool, expression are well established, we have focused our efforts on United Kingdom; 2University of Warwick, Department of Statistics, the development of a method for semi-quantitative determination Warwick, United Kingdom; 3University of Warwick, Warwick of the genome-scale genetic interactions in a high-throughput Systems Biology Centre, Warwick, United Kingdom; 4University of manner. We have shown on the rapamycin example that it is Manchester, School of Medicine, Manchester, United Kingdom; possible to accurately identify the nodes (genes/proteins) of 5University of Edinburgh, Queen’s Medical Research Institute, the molecular network perturbed by a given biologically active Edinburgh, United Kingdom; 6University of Manchester, School of compound from such experimental data. Medicine, Manchester, United Kingdom Conclusions: To determine the type and direction of all interactions between the nodes, highly accurate quantitative Objective: Cells and tissues have to regulate the dynamics phenotypic data is required. Yeast as a unicellular model of transcription and handle the uncertainty and potential noise organism is mostly valuable here, since a cumulative phenotype, that results from there being only two copies of each gene. The which reflects the overall physiological state of the cells, can be previously observed instability in activity of the prolactin promoter determined accurately and in a medium-throughput manner by may be intrinsic to promoter function, reflecting stochastic measuring growth curves. We conclude that the combination variability, or due to ‘extrinsic’ cellular factors. We therefore of gene expression and genetic interaction data, obtained and generated dual-transfectant cell lines, where the hPRL 5000bp analyzed by the state-of-the-art experimental and bioinformatics promoter directed luciferase and destabilised EGFP (d2EGFP) approaches, is an extremely valuable tool for biological discovery. expression, to allow independent monitoring of both reporter This work was supported by grants J2-9699 and J1-6507 from genes simultaneously in real time. the Slovenian Research Agency. Results: Live-cell imaging of luminescent reporter gene expression revealed large quantitative changes in prolactin promoter activity in single clonal GH3 rat pituitary cells (stably expressing luciferase under the control of the human prolactin (hPRL) 5000bp exon 1b promoter sequence), and single transgenic rat pituitary cells (expressing a 160kb hPRL BAC construct containing the firefly luciferase gene). Prolonged parallel live-cell imaging of both fluorescence and luminescence showed that each reporter within individual cells displayed fluctuating expression that varied between cells. These fluctuations did not appear to be correlated between the reporter genes. Mathematical reconstruction (MCMC) of transcription rates showed that this variability was attributable to independently

104 ICSB 2008 varying rates of transcription within single cells and suggested requires a special attention. We started from a simplified picture: that cycles of gene transcription had a regular period of 7-12 mitogens activate CycD1/CDK complexes which phosphorylate hours. The two genes were however frequently out-of-phase. RB, which in turn release the hold on E2F transcription factors The correlation between dual reporter expression in single cells that control cell cycle progression. We added several tens of increased following trichostatin A-treatment, suggesting that proteins involved in cell cycle regulation, carefully connecting intrinsic variability between prolactin gene loci may depend on them to the rest of the network by means of known biochemical chromatin status. transformations. The resulting map has a total of 78 proteins Conclusions: We report a novel approach to reconstruct represented in 215 chemical species, 530 reactions 176 transcription dynamics from single cell imaging data. Dynamic genes, and compiles experimental results from more than 350 pituitary gene transcription is attributable, at least in part, to publications. The interactive RB pathway map is available at intrinsic promoter noise and oscillating transcription and is http://bioinfo.curie.fr/projects/rbpathway/interactive/rb_network. regulated by chromatin conformation. html. Objective: One purpose of the construction of this diagram, DS2-1-56 once validated, is to provide a map of RB pathway that can become not only a reference when studying different cancers and Functional, spatial and temporal characterization of mutations but also a tool to analyze formally the pathway and Mycoplasma pneumoniae transcriptome predict its behavior in response to different deregulations. Güell, Marc1; Yus, Eva1; Gavin, Anne-Claude2; Serrano, Luis1 Results: To use the diagram as a modeling tool, we needed a 1Centre for Genomic Regulation, EMBL-CRG Systems Biology more abstract view of the network. Thus, we decomposed RB Unit, Barcelona, Spain; 2European Molecular Biology Laboratory, pathway into 24 network modules using automatic techniques Structural and Computational Biology Unit, Heidelberg, Germany and consequent manual curation. A Cytoscape BiNoM plugin (http://bioinfo.curie.fr/projects/binom/) was developed to perform Objective:With only 689 genes Mycoplasma pneumoniae is such analysis. We then integrated CGH and transcriptome among the simplest known organisms. Because of its simplicity, bladder cancer data in our diagram and were able to verify and mycoplasma represents an attractive organism for systems- identify which group of proteins (modules) were upregulated or wide analysis. Such approaches aiming at the whole quantitative downregulated as a function of cancer progression. Interestingly, understanding of an entire organism are expected to guide the classification of the samples revealed two groups of tumors rational engineering/modifications. Several comprehensive which closely matched two different progression pathways datasets are being collected (structural characterization,protein previously identified in bladder cancer (based on anatomoclinical complexes, metabolic flows) that will ultimately be integrated in data and molecular parameters). a coherent model using the algorithm SmartCell(http://smartcell. Conclusions: Analyzing real data with our map confirmed known Posters crg.es). In this abstract, we present one of these datasets: results and showed new interesting ones. Dedicated the functional, spatial and temporal characterization of M. pneumoniae. DS2-1-58 Results:Three different DNA chip technologies have been used in order to map different aspects of M. pneumoniae transcriptome. The mathematics of tanning Transcriptome functional organization was addressed using a Thingnes, Josef1; Oyehaug, Leiv2; Hovig, Eivind3; Omholt, Stig4 custom DNA array. Up to 94 chips have been used to measure 1CIGENE, Ås, Norway; 2CIGENE and Norwegian University of Life mRNA levels under different conditions (stress, synchronization, Sciences, Department of Mathematical Sciences and Technology, different carbon sources...). The data has been normalized and Ås, Norway; 3The Norwegian Radium Hospital, University of clustering methods have been used so as to classify samples and Oslo, Department of Tumour Biology, Oslo, Norway; 4CIGENE, genes. Different gene modules have been identified to respond Norwegian University of Life Sciences, Department of Animal and to different stresses. A whole transcriptome mapping has been Aquacultural Sciences, Ås, Norway carried out using an Affymetrix tiling array that interrogates both strands of the genome at a resolution of 8 bases. From Objective: The pigment melanin is produced by specialized cells this technique we obtain the complete set of transcripts and called melanocytes. In healthy skin, melanocytes are sparsely its relative abundance. This not only helps to define the exact spread among the other cell types in the basal layer of the location, including transcription start and termination of the of epidermis. Sun tanning results from an Ultra Violet (UV) induced protein coding genes, but is also a powerful tool for de novo RNA increase in the release of melanin to neighbouring keratinocytes, discovery. The temporal behaviour of the transcripts is studied the major cell type component of the epidermis. We still have a using an specially designed Agilent array. This chip targets every modest understanding of the tanning response. gene and allows radioactivity labeling.Thus, enabling transcripts Here we provide a mathematical model of the tanning response half-life measurements. describing the intercellular dynamics between melanocytes and Conclusions: We have established a high resolution global map keratinocytes following UV exposure. In healthy skin, melanocyte of M. pneumoniae transcriptome and infered the gene regulatory growth is strictly controlled by surrounding keratinocytes. network. Despite its simplicity, it is clear yet that this small Melanocytes that develop into malignant melanoma escape this bacterium is capable of regulating gene expression dynamically in control. A better grasp of the regulatory communication between response to environmental perturbations. these two cell types in connection with the natural tanning response may thus be instrumental for better understanding of DS2-1-57 the induction of malign melanoma. Results: Assuming available experimental data, the model offers Module deregulations of RB/E2F pathway in cancer an opportunity to mimic UV exposure of different intensities and Calzone, Laurence1; Gelay, Amélie2; Zinovyev, Andrei3; Radvanyi, duration on different skin colours and photo types. In simulation, François4; Barillot, Emmanuel3 the model describes quite accurately the tanning response 1Institut Curie, Service Bioinformatique, INSERM U900, Paris, of six out of nine individuals of different skin types, but fails to France; 2Institut Curie, Centre de Recherche, Paris, France; describe the three remaining. The epidermal layer thicknesses 3Institut Curie, Service Bioinformatique, INSERM, U900, Paris, and the maximum length of melanocyte dendrites that develop as France; 4Institut Curie, CNRS UMR144, Institut Curie, Centre de response to UV exposure, differ according to our model between Recherche, Paris, France individuals indicating the importance of these quantities for the tanning response. RB (RB1) is a tumor suppressor gene involved in many cancers, Conclusions: The fact that the model fails to describe parts in both familial and sporadic forms, either through inactivating of the data emphasises that the current understanding of the genetic and epigenetic mutations of RB gene or as a result of a tanning physiology, on which the premises of the model are deregulation of the kinases that control its activity. Because of based, can be improved. Thus the model’s ability as well as its implication in so many cancers, the study of RB regulation its failure to describe empirical data can be seen as assets in

ICSB 2008 105 functioning as guideline for future experimental programmes. used to link phosphoproteins to cytokine release. Two step MLR was implemented to link cues+inhibitors to signals and DS2-1-59 cytokine release. Transformation patterns among normal and cancer hepatocytes were ranked based on the differences of their FAR1 gene dosage affects metabolism and ribosome correlations values. To visualize differences between cell types, biosynthesis MLR connectivities were overlaid on a literature-curated pathway Vanoni, Marco; Querin, Lorenzo; Alberghina, Lilia map where the thickness of the lines correlated with the strength Università di Milano-Bicocca, Biotecnologie e Bioscienze, Milano, of the MLR value. Italy Conclusions: Pathway maps were considerably different and many potential tumorgenic signatures were revealed on HCCs: Objective: Nutrients are the main environmental determinants, downregulation of the stress related pathway, upregulation of beside mating factors, that affect cell cycle progression in pro-survival protein activities, engagement of angiogenesis, budding yeast. Two quantitative parameters characterize each sensitivity to pro-growth stimuli. Evasion of immune surveillance exponentially growing population: the rate of growth (λ, min- via an altered NFκB pathway seems to be a characteristic to all 1) and the critical cell size at the entrance into S phase (Ps). HCC cell lines that confer a strong survival advantage for tumor Higher growth rates and larger Ps are observed in rich media. progression. Our laboratory presented a mathematical model of the G1 to S network that newly takes into account nucleo/cytoplasmic DS2-1-61 localization, the role of the cyclin-dependent kinase Sic1 in facilitating nuclear import of its cognate Cdk1-Clb5, Whi5 control, A robust mathematical model of the calcium calcium and carbon source regulation of Sic1 and Sic1-containing dependent release and Its regulation by CaMK in a cardiac complexes and the role of Far1 in inhibiting the Cln3/Cdk1 myocyte complex (Barberis et al., PLoS Comp Biol 2007, e63). In order Koivumäki, Jussi1; Takalo, Jouni1; Korhonen, Topi2; Weckström, to better understand the role of Far1 in cell physiology, in this Matti1; Tavi, Pasi2 report we investigate the changes in transcriptome and proteome 1University of Oulu, Department of Physical Sciences, Oulu, induced by changing FAR1 dosage. Finland; 2University of Oulu, Department of Physiology, Oulu, Dedicated Results: far1 and FAR1-overexpressing strains were analyzed Finland

Posters Δ during exponential growth in synthetic complete medium supplemented with either 2% ethanol (SCE) or 2% glucose (SCD) Objective: In cardiomyocytes, the calcium flux through the using both transcriptomics and proteomics techniques. Our release channels (Ryanodine Receptors; RyRs) defines, on a results indicate that FAR1 overexpression affects transcription beat-to-beat basis, transient changes of intracellular calcium and translation of gene products involved in cell metabolism corresponding to the trigger, i.e., the transmembrane calcium (both in glucose and in ethanol), increases ribosomal protein current, and therefore determines the dynamics of the contractile translation/stability (in glucose) and, finally, brings to increased function of myocytes. A controversy exists related to the exact RNA accumulation. mechanisms responsible for the activation and termination of Conclusions: Reported results above are in line with ribosome calcium release, as well as to the increased leakiness of RyR, biogenesis being a central element in the nutrient-dependent which has been reported in association with various heart failure control of cell cycle progression and reveal an unexpected conditions, thought to be due to excess phosphorylation by link between Far1, a dosage-dependent inhibitor of the G1/S calcium/calmodulin dependent kinase II (CaMK). Our aim was to transition, ribosome biogenesis and metabolism. develop a mathematical model of RyR that would describe both the dynamics and the CaMK-dependence of the calcium release. DS2-1-60 Results: During the last two decades, a number of different RyR models have been published, taking either the deterministic or Unraveling transformation patterns in cancer hepatocytes the stochastic approach. The former has obvious advantages using high-throughput protein activity-based profiling in relation to the computational cost and thus enables the RyR Alexopoulos, Leonidas1; Saez-Rodriguez, Julio2; Cosgrove, model to be implemented into a comprehensive ventricular Benjamin3; Lauffenburger, Douglas3; Sorger, Peter4 myocyte model. Based on previously published models, we 1MIT & National Technical University of Athens, Mechanical present a deterministic model of RyR that includes CaMK- Engineering, Zografou Campus, Athens, Greece; 2Harvard dependent regulation of the release. We validate the model Medical School and MIT, Systems Biology, Boston, United States; by comparing the simulated results to corresponding in vivo 3MIT, Biological Engineering, Cambridge, United States; 4Harvard observations, both under normal and CaMK overexpression Medical School, Systems Biology, Boston, United States conditions. Results show that the model describes, in a quantitative way, the functional connection between the calcium Objective: To determine signaling events that distinguish normal trigger and release, the dynamic regulation by CaMK, as well as from cancer human hepatocytes. Normal hepatocytes and the role of RyR as a component of the whole myocyte. 4 hepatocellular carcinoma (HCC) cell lines (HepG2, Hep3B, Conclusions: Based on the results we conclude that the Huh7, and Focus) were utilized. A cue-signal-response (CSR) presented model is robust and computationally efficient protein-based dataset was created that covers a wide range of description of the function of RyR. The model is a solid hepatocyte phenotype. ~26,000 protein measurements were component that can be employed in further modelling studies, made under 88 different perturbations generated by orthogonal e.g., to explore the double-edged role of CaMK in various heart co-treatments with a diverse set of ligands and inhibitors. As failure conditions. pro-inflammatory and acute phase stimuli we chose TNFα, IL1α, IL6, and INFγ; for innate immunity we chose LPS; for DS2-1-62 the insulin pathway we chose IGF-I; for pro-growth signals we chose TGFα. For each stimulus, 7 inhibitors were chosen that A model of dendritic cell maturation: Using Interleukin 12 target 5 pathways (MEK, p38, NFκB, Akt, and JNK). For each as a model cytokine cue+inhibitor perturbation, 17 intracellular phosphoproteins and Gonzalez-Lergier, Joanna; Sealfon, Stuart 50 extracellular cytokines were collected at three time points. The Mount Sinai School of Medicine, New York, NY, United States dataset was created using a high-throughput method of bead- based fluorescent readings (Luminex). Assays were optimized for Dendritic cells are antigen-presenting cells responsible for the multiplexability and checked for donor-to-donor and preparation- initiation of immunity. Immature DCs act as sentinels that, upon to-preparation variability as well as purity from nonparenchymal recognizing a pathogen, alter their gene expression pattern; the liver cells. mature DCs then secrete signals to stimulate the appropriate Results: Multi-linear regression (MLR) and Boolean models immune response. One of these signals is the cytokine IL-12, were used to analyze the dataset. Single step MLR model was composed of IL-12p35 and IL-12p40; the interaction of both

106 ICSB 2008 subunits is required for the production of bioactive IL-12. A DS2-1-64 related cytokine, IL-23, is formed by the interaction of IL-12p40 with IL-23p19. The relative proportions of IL-23 and IL-12 depend Dynamic modeling of phototransduction biochemistry in on the pathogen recognized by the DC. Consequently, a model vertebrate rods: From dark/light adaptation to disease is being built in order to attain a more accurate understanding Dell’Orco, Daniele1; Fanelli, Francesca1; Schmidt, Henning2 of IL-12 synthesis. The first part of the model encompasses 1University of Modena and Reggio Emilia, Dept. of Chemistry and the transcriptional activation of IL-12 genes. Previous research Dulbecco Telethon Institute, Modena, Italy; 2University of Rostock, has identified some of the transcription factors involved and it Rostock, Germany has been found that the transcriptional regulation of IL-12 and IL-23 involves both common and distinct signaling pathways. Objective: The rhodopsin-mediated signal transduction However, further experiments are required to obtain a more pathway involved in vertebrate phototransduction contains an complete signaling network from virus infection to the activation extremely efficient amplification cascade. Even the absorption of these transcription factors. The second part of the IL-12 model of a single photon by a single photoreceptor cell leads to involves the synthesis of bioactive IL-12 and mass-action kinetic a detectable electrical signal in the retina. In this work, we equations have been developed to approximate the transcription present a quantitative dynamic model describing the underlying and translation of the two subunits, their dimerization and biochemistry and molecular events. It is based on a previously secretion. Experimental data are currently being obtained for published model that has been significantly extended towards kinetic parameter estimation, as well as for testing the model; a finer description of the shutoff-regulatory mechanisms of both this proposed methodology can then be extended to other receptor (rhodopsin) and effector (PDE-Gα-GTP) molecules. DC maturation signals in order to acquire a more complete The ultimate goal of the study is to achieve, for the first time, understanding of the initiation of immunity. a computational model able to probe hypotheses on the dark and light adaptation mechanisms potentially associated with the DS2-1-63 spontaneous activity of the opsin in the absence of light stimulus. The latter condition is found in several forms of genetic retinal Transcriptional profiling of normal and transformed diseases such as Leber congenital amaurosis and congenital fibroblasts: Effect of growth under limiting glucose night blindness. concentration Results: The model has been successfully simulated in both Balestrieri, Chiara1; Chiaradonna, Ferdinando1; Liberati, Diego2; stochastic and deterministic frameworks in order to reproduce Vanoni, Marco1; Alberghina, Lilia1 single photon responses and outputs from higher-intensity 1Università di Milano Bicocca, Dipartimento di Biotecnologie e illumination experiments, respectively. Furthermore, the 2 Bioscienze, Milano, Italy; Consiglio Nazionale delle Ricerche simulations show that the model is capable of reproducing Posters Politecnico, Milano, Italy a broad range of experimental in vivo conditions, including Dedicated wild type and genetically manipulated rod cells, in which Objective:The link between cancer genetics and abnormal the photoresponses were measured by electrophysiological use of glucose by tumors is opening a new scenario in which techniques to probe dark and light adaptation mechanisms. bioenergetics would contribute to, and sustain, malignant The variety of experimental data available concerning both the transformation. We showed that K-ras transformed mouse biochemistry of the pathway and the physical setting of the stimuli fibroblasts are exquisitely sensitive to glucose deprivation, imposes tight modeling constraints regarding the parameterization show reduced oxidative phosphorylation ability and enhanced of the model’s kinetic expressions. apoptosis under glucose depletion as compared to their normal Conclusions: The developed model constitutes a valuable counterparts. Ras-dependance of the observed phenotypes was tool to investigate visual transduction dynamics in normal and confirmed by analysis of transformed cells that expressed a Ras- disease-associated conditions. Furthermore, it might represent a inhibitory dominant negative GEF. The goal of the present study robust paradigm for other G-Protein Coupled Receptor-mediated was to evaluate time-course changes in transcriptional profiles of signaling pathways. normal and transformed cells grown under different initial glucose concentrations in order to identify potential biological pathways DS2-1-65 involved in such metabolic rerouting associated to transformation. Results: Gene Ontology (GO) annotations indicate that the Diffusion Limited Space (DLS) and its impact on water and majority of regulated genes belong to Metabolism, Transport, Cell K+ homeostasis in kidney principal cells and astrocytes organization and biogenesis, Cell cycle and Cell communication/ Kamali-Zare, Padideh1; Kowalewski, Jacob M.1; Aperia, Anita2; Signal transduction. Different tools (Database for Annotation, Brismar, Hjalmar1 Visualization, and Integrated Discovery –DAVID- and GeneSpring) 1Royal Institute of Technology, Cell Physics, Applied Physics, were used to map differentially expressed genes to known Stockholm, Sweden; 2Karolinska Institutet, Woman and Child biological and physiological pathways (KEGG). Current work is Health, Stockholm, Sweden focused on increasing precision of the obtained results by use of other methods of analysis. We are also working on integration Objective: Diffusion Limited Space (DLS) is a part of the of gene expression data with other relevant parameters, such as extracellular space where diffusion is limited by the shape and protein interaction maps, as a first step to construct an interaction size of the region. DLS can be formed by the morphology of map able to evidence glucose-modulated pathways in normal the cell itself or by a close proximity of two cells placed next to and transformed cells. each other. The synaptic region in brain, intercellular spaces in Conclusions: The medium-term aim of this project is to kidney and the membrane foldings in kidney principal cells are reach a better understanding of networks originating fragility examples of DLS. The importance of these cells for K+ transport of transformed cells in response to glucose shortage. In in a steady state and the localization of water and K+ channels in perspective, this knowledge will be valuable for rational, system- the membranes facing DLS, raise the question whether DLS can level identification of therapeutic targets for improved anti-cancer have a role in the function of proteins localized in DLS. In order to drugs. address this question we have formulated the key mechanisms for DLS K+ and water transport analytically and applied them into a series of biophysical models of kidney principal cells and astrocytes (using Virtual Cell: www.vcell.org), where we have tested and clarified our hypothesis applying different intra- and extracellular geometries together with different sets of components and boundary conditions. Results: The models show that the presence of a DLS around K+ and water channels creates an efficient interaction between K+ and water, resulting in a fast and passive K+ homeostasis in

ICSB 2008 107 DLS, important for K+ shunting via Kir-channels (Kir 7.1) in kidney (www.vcell.org) [2]. Our model contains geometrically polarized principal cells, K+ siphoning via Kir 4.1 in astrocytes and volume regions with increased levels of Kir channels and aquaporin-4 regulation of both cell types. Our results also indicate that DLS (AQP4) water channels, known to be co-expressed with Kir regulating the function of DLS-AQP and Kir channels can lead to channels [3]. maintenance of the cell membrane potential, important for the net Our model contains an astrocyte surrounded by EC regions of K+ uptake by the cell in a steady state. different geometrical sizes between which diffusion is limited. Conclusion: Here we have developed a non-classical modeling Computer simulations are used to study the effect of K+ and method by combining analytical and biophysical modeling water channels in this system. together with an extra dimension arising from the interactive Results: Our simulations show that Kir channels can cause system involving major components. This method can be applied K+ current from the narrow EC space in the synaptic region to those biological questions that very little are known about through the astrocyte endfoot and into a larger EC region. The them. rapid decrease of K+ in the narrow EC region causes decreased osmolarity and shrinkage of that region by water uptake through DS2-1-66 AQP4. Our results show that water follows K+ into the astrocyte and that water flux through the endfoot membrane is bidirectional. Nutritional control of the G1 to S transition: Steps towards Conclusions: Our quantitative data shows that coexpression of network identification and parameter estimation Kir and AQP4 can direct K+ and water fluxes to certain domains Coccetti, Paola1; Cirulli, Claudia1; Tripodi, Farida1; Busti, Stefano1; of astrocyte membrane. Our results also indicate that the AQP4 Gotti, Laura1; Tsiarentsyeva, Viktoryia1; Samalikova, Maria1; enriched endfoot membrane of an astrocyte can protect narrow Grandori, Rita1; Hohmann, Stefan2; Vanoni, Marco1; Alberghina, EC regions from uncontrolled volume changes. We conclude that Lilia1 AQP4 can enhance K+ spatial buffering by astrocytes and that 1University of Milano Bicocca, Italy, Dipartimento di Biotecnologie regional distribution of AQP4 is important for controlling volume e Bioscienze, Milano, Italy; 2Goteborg Univerity, Goteborg, changes in astrocytes and their surrounding EC space. Sweden References: [1] Orkand, R.K., et al. Neurophysiol, 1966. 29(4): p. 788. Objectives: Our group is developing a modular systems biology [2] Slepchenko, B.M et al. Annu. Rev. Biophys. Biomol. Struct., Dedicated approach to yeast cell cycle that relies on molecular analysis 2002. 31(1): p. 423. Posters of perturbed cell cycle for network identification and parameter [3] Nagelhus, E.A., T.M. Mathiisen, and O.P. Ottersen. estimation. Control of events of the G1 to S transition by nutrients Neuroscience, 2004. 129(4): p. 905. (notably glucose) and phosphorylation-dependent intracellular pathways (notably Snf1 and CK2 pathways) could offer a way DS2-1-68 to refine the network structure of the G1 to S transition (Barberis et al., PLoS Comp Biol 2007, e63) and its connection to major Differentiating between Hog pathway dependency in signalling pathways. translational and transcriptional adaptation to saline stress Results: The signalling and metabolic role of glucose in in yeast modulating the G1/S transition is being investigated by studying Warringer, Jonas cell growth/cell cycle coordination in mutant strains unable to University of Gothenburg, Gothenburg, Sweden transport glucose or defective in glucose sensing. The study of these mutants under steady state and transient growth conditions Yeast physiological adaptation to an external increase in Na+ is (i.e., ethanol to glucose nutritional shift-up) indicates that control one of the best characterized cellular stress responses but has by glucose on cell growth/cell cycle coordination is distributed mostly been studied on the level of transcriptional regulation. between sensing and metabolism. Snf1-mediated signal Here we perform a time-resolved differentiation between the transduction pathway affects the level and subcellular localization translational and transcriptional response to elevated Na+ in of key cell cycle proteins as a function of growth conditions and yeast, both in the presence and absence of a functional Hog cell cycle phases. The molecular basis of these control is under (High Osmolarity Glycerol) pathway. We find that the translational investigation. The role of CK2 phosphorylation on the G1/S salt-response precedes the transcriptional response, peaking at transition is approached by analysing the phosphorylation state 6 minutes after stress addition, suggesting that the earliest salt of relevant players (i.e. Sic1, Cdc34) as a function of growth response utilizes already existing transcripts. A large sub-fraction conditions and cell cycle position. Our results suggest that CK2 of transcriptionally induced mRNA are either not further mobilised can effectively integrate signals required to modulate the onset or signficantly down-regulated on a translational level. Similarly, a of S phase by phosphorylating multiple substrates involved in the smaller fraction of genes are salt induced on a translational level G1/S transition. but not in terms of transcription. Particularly interesting is a strong Conclusions: Experiments described herein fall within a large translational salt induction of most low affinity hexose transporters concerted experimental effort of our laboratory integrated in the at the earliest time-points, suggesting an initial elevation of UNICELLSYS FP7 project, in which biochemical information glucose import. With the notable exception of ion transport and is iteratively used for network identification, refinement and glycerol accumulation few of the dominant functional trends of parameter estimation in order to construct a large scale translational salt induction were dependent on a fully functional mathematical model of cell cycle in yeast. Hog pathway. Our findings highlight the differences between translational and transcriptional responses to fluctuations in the DS2-1-67 external milieu and underscore the limitations of focusing to narrowly on transcriptional events. Modeling the role of Aquaporin-4 in astrocyte K+ spatial buffering DS2-1-69 Kamali-Zare, Padideh; Kowalewski, Jacob M.; Brismar, Hjalmar Royal Institute of Technology, Cell physics, Applied physics, Presentation of SITCON project: Modeling signal Stockholm, Sweden transduction induced by a chimeric oncogene Stoll, Gautier1; SITCON, Consortium2 Objective: Neuronal activity causes K+ efflux leading to increased 1Institut Curie, service bioinformatique, Paris, France; 2Institut K+ levels in the narrow extracellular (EC) space in between Curie, Institute of Mathematical Research of Rennes, IRISA, neurons and astrocytes in the brain. Astrocytes express inward INRIA, http://bioinfo.curie.fr/projects/sitcon/, Paris, Rennes, rectifying K+ (Kir) channels that are involved in transporting K+ from France regions of high concentration to regions with lower K+ levels, a mechanism known as K+ spatial buffering [1]. We have created Objective: In Ewing tumor, most of the cells have the same a spatial computational model to simulate this mechanism via genetic modification: a translocation producing the chimeric gene electrochemical driving forces in an astrocyte using Virtual Cell EWS/FLI-1. We present the SITCON project, whose goal is the

108 ICSB 2008 understanding of EWS/FLI-1 effect on tumor phenotype. This identification of critical signaling components and potential targets work is a intense collaboration between different research groups for cancer therapeutics. of experimentalists and theoricists. Results: We based our approach on different types of data: DS2-1-71 transcriptome time series of EWS/FLI-1 inducible cell lines, micro-RNA time series of EWS/FLI-1 inducible cell lines, patient Molecular and computational analysis of regulation of transcriptome, patients CGH, patient microRNA profiles. Among hSos1, the major activator of the proto-oncoprotein Ras the different data analysis, we present two types of interesting Sacco, Elena1; Farina, Marcello2; Greco, Claudio1; Busti, Stefano1; intermediate results: DeGioia, Luca1; Fantinato, Sonia1; Liberati, Diego3; Alberghina, 1) Analysis of transcriptome time series combined with literature Lilia1; Vanoni, Marco1 data mining produced an annotated influence network. The 1Università di Milano Bicocca, Dipartimento di Biotecnologie e consistency between the data and the network has been tested, Bioscienze, Milano, Italy; 2Politecnico di Milano, Dipartimento di using simple pathway consideration and qualitative dynamical Elettronica e Informazione, Milano, Italy; 3Consiglio Nazionale delle modeling. Experimental investigations is under process; we will Ricerche Politecnico, Milano, Italy confront the network but also identify important genes/proteins involved in the tumor progression. Objective: Sos proteins are Ras-specific Guanine nucleotide 2) Correlation analysis of transcriptome and microRNA profiles Exchange Factors (GEF). Activation of Ras-specific GEF activity in patient, combined with sequence analysis, produced pairs of requires growth factor-dependent recruitment of hSos1 to the (likely interacting) microRNA-mRNA that are involved in Ewing plasma membrane via Grb2 and intra-molecular rearrangements cancer phenotype. The integration with genes selected by in order to release the Ras-GEF domain from inhibitory constraints transcriptome time series analysis led to several modules of exerted by its flanking amino- and carboxyl-terminal regions. key genes in EWS/FLI-1 system that are regulated by the same Structural determinants of the Sos auto-inhibitory mechanisms microRNA. are not yet entirely understood. Conclusion: A global picture of EWS/FLI-1 system starts Results: The major elements of novelty derived from our analysis to emerge from our intermediate results. In particular, the of Sos regulation are the following. Down-regulation of Sos influence network is the basis for different mathematical models, Ras-GEF activity requires both the DH and HF domains whose including tumor evolution. SITCON structure is based on a close inhibitory function may be regulated through their competition collaboration between experimental and theoretical groups, which for binding to PH. As a result, the amino-terminal region exhibits forms the ground for building a global theoretical model of Ewing alternative structural arrangements with different regulatory tumor. competence on the Ras-GEF activity of Sos. The inhibitory

activity exerted by the HF-DH domains is reinforced by their direct Posters DS2-1-70 interaction with the carboxyl-tail whose major inhibitory effect Dedicated on Ras-GEF activity is to stabilize the amino-terminal domain Bayesian network inference models for identification of of Sos in a non-catalytically competent state towards Ras. A new components of RAS-MEK-ERK signaling network mathematical model was developed according to above data Gendelman, Rina1; Xing, Heming2; Bayani, Nora3; Feiler, Heidi3; and the resulting dynamics of Sos inter-domain interactions were Gray, Joe3; Khalil, Iya2; Korn, W. Michael1; Mirzoeva, Olga1 simulated. In all cases studied in-silico, steady state simulated 1University of California San Francisco, Medicine, San Francisco, complementation ability was found to closely match experimental United States; 2Gene Network Sciences, Boston, United States; findings obtained with Sos deletion mutatants. Sensitivity analysis 3Lawrence Berkeley National Laboratory, Berkeley, United States indicates that our mathematical model not only is able to describe the behavior of Sos deletion, but modulation of activity of the full Development of new cancer therapies relies heavily on our insight length molecule as well. into signaling networks underlying cancer progression. Although Conclusions: Within a comprehensive systems biology- the RAS-MEK-ERK is one of the most studied networks, our generated model of RTKs signalling pathway(s), Sos is located understanding of it is still incomplete. We are aiming at expanding near the core. Here we “zoom in” on such a central node and knowledge about such signal transduction networks by using a report that the activation state of hSos1, arises as an “emergent Bayesian Network Inference approach. property” of the multi-domain structure of this complex protein, A model of gene interactions in breast cancer cells treated with thus allowing multi-level integration of a complex network of intra- MEK inhibitor U0126 was developed. This model related gene and inter-molecular signals. expression profiles to the physiological end-point of G1 arrest. The model then was used to identify genes that have the most DS2-1-72 impact on the cells in context of RAS-MEK-ERK signaling and cell cycle progression. Each gene in the model was subject to Systems biology approach to quantify signalling pathways forward simulation such as in silico knockdown (KD) or over- in B cell chronic lymphocytic Leukemia expression experiments. The effects on cell cycle were ranked Bhattacharya, Nupur1; Caudron, Maiwen2; Szabowski, Axel3; with Kolmogorov-Smirnov statistic accordingly. The in silico Busch, Hauke4; Lichter, Peter5; Stilgenbauer, Stephan1; Rippe, experiments revealed three genes whose KD had the highest Karsten2; Mertens, Daniel1 effect on G1 phase - BTG3, TRIB1, and C14ORF133. Our model 1University of Ulm, Internal Medicine III, Ulm, Germany; predicted that KD of BTG3 and TRIB1 will result in increase in 2BIOQUANT & DKFZ, ”Genome Organisation and Function”, G1 phase of the cell cycle. Interestingly, TRIB1 has been recently Heidelberg, Germany; 3DKFZ, ”Signal Transduction”, Heidelberg, identified as a binding partner of MEK1. A second model was Germany; 4DKFZ, ”Theoretical Bioinformatics”, Heidelberg, obtained by addition of 22 known cell cycle regulators to the Germany; 5DKFZ, ”Molecular Genetics”, Heidelberg, Germany original model. Model-2 was subject to the same simulation protocol as previous model. Although new genes were added to Objective :CLL is characterized by an accumulation of malignant the model, predictions still contained the three genes identified by B cells. While CLL cells survive in-vivo, they undergo rapid the first model. In addition, CCND1 and SOD1 were predicted to apoptosis when cultured in-vitro. This highlights the importance significantly impact G1 status. Using siRNA against TRIB1, BTG3, of the microenvironment for survival of CLL cells. Several signal C14ORF, CCND1 and EGR1 and SPRED2 we are validating transduction pathways involved in survival of CLL cells have the model predictions in vitro. As predicted, our preliminary been studied, but so far only qualitatively. Moreover, it is largely data demonstrate that KD CCND1 and TRB1 lead to significant unknown how these signalling pathways are interconnected to increase in G1. KD of EGR1 and SPRED2 did not have any effect result in enhanced survival. The objective of this project is to on the cell cycle as expected. Treatment of cells with BTG3 siRNA model signalling pathways in CLL by culturing primary CLL cells leads to a small but not significant increase in G1. with several external stimuli that will be systematically varied Our results show that data-driven simulation combined with in intensity (e.g. ligands, serum, co-culture with stromal cell experimental validation might be a powerful approach for lines). Subsequently, phenotypic parameters like changes in

ICSB 2008 109 transcriptome and survival rates will be quantified by fluorescence for causal interpretation of gene expression data. It performs a microscopy and microarray kinetics. Results obtained from rather unusual way of analysis through considering the earlier these quantitations will be combined into a coarse-grained and causes that have led to the observed gene expression changes continuous time recurrent neural network (CTRNN) modelling. rather then analysing the later effects of those changes. First Results:We incubated primary CLL cells with CD40L, IL4, SDF-1, of all, promoters of differentially expressed genes are analyzed BCR, APRIL, BAFF, anti-IgM and combinations thereof. Apoptosis and specific combinations of transcription factors (Composite and proliferation was quantitated using four-colour flow cytometry. Modules) regulating these genes are hypothesized. Next, analysis In preliminary experiments, spatial distribution of CD40 receptor of signal transduction network upstream of these transcription (CD40R) and IL4 receptor was compared between CLL cells and factors allows us to reveal key signaling molecules that can non-malignant cells using confocal microscopy in order to show master the observed gene expression profile. The method utilizes transformation-associated deviations. data from three databases (TRANSFAC®, TRANSPATH® and Conclusions:We identified IL4 and CD40L as ligands that HumanPSD http://www.biobase-international.com/ ). induce the strongest response in primary CLL cells. Even though Affymetrix microarray data have been taken from time series absolute values of apoptosis and proliferation were different study of retinoic acid induced differentiation of promyelocytes to among patients, the resulting phenotypic patterns are the same neutrophils. Sets of promoters of differentially expressed genes in several patients. In addition, CD40R molecules clustered on on different time points have been compared to promoters of CLL cells, which points to differential signalling in CLL cells as genes that did not showed any significant change of expression. compared to non-malignant cells. Also in contrast to healthy We revealed highly significant combinations of transcription factor B cells, we see nuclear localization of CD40R in CLL cells. In binding sites for such factors as Egr-2, Myc, SOX-6, IRF-1, PAX- summary, CD40 and IL4 comprise promising molecules in the 3, GATA and C/EBP. Finally, the analysis of the signal transduction survival of CLL cells. pathways upstream of these transcription factors helps to identify several potential key molecules such as calcineurin and PKAc. DS2-1-73 Conclusions: This analysis helps to build a network model of neutrophils differentiation enabling generation of hypothesis on Systems biology of osteoblast differentiation dynamic mechanisms of cell differentiation. Moné, Martijn J.1; Bruggeman, Frank J.2; Eijken, Marco3; van Parts of the work were funded by EU grants “TRANSISTOR” Dedicated Leeuwen, Johannes P.T.M.3; Westerhoff, Hans V.1 (MRTN-CT-2004-512285), “EuroDia” (LSHM-CT-2006-518153), Posters 1Vrije Universiteit Amsterdam, Molecular Cell Physiology, VALAPODYN (LSHG-CT-2006-037277) and Net2Drug (LSHB- Amsterdam, Netherlands; 2Netherlands Institute for Systems CT-2007-037590). Biology, Multiscale Modelling and Nonlinear Dynamics, Amsterdam, Netherlands; 3Erasmus MC, Internal Medicine, DS2-1-75 Rotterdam, Netherlands Specificity in cell signaling: Experiment and theory Normal cell physiology is controlled by a multitude of intertwined Bardwell, Lee cellular processes. Perturbations herein can give rise to cellular University of California, Irvine, Developmental & Cell Biology, dysfunction and, hence, to disease. Osteoporosis is a complex Irvine, United States multi-factorial disorder in which bones exhibit reduced mineral density. Correct and timely osteoblast differentiation plays an Objective: Signal transduction pathways are often important role in the control and maintenance of bone quality and interconnected to form larger networks. One pathway may cross- several growth factors and hormones affect the differentiation regulate another pathway, and several distinct pathways may process and the formation of an extracellular protein matrix share components. Such cross-regulation may enable the cell with proper mineralizing capacity. Mesenchymal stem cells can, to integrate its overall response when receiving multiple stimuli. for instance, be driven to differentiate into osteoblasts upon However, extensive interconnections increase the difficulty of stimulation with epidermal growth factor. Yet also the bone- maintaining specificity from signal to cellular response; that is, forming capacity of lineage-committed pre-osteoblasts can still they increase the likelihood that (under certain circumstances) be tuned; e.g. activin A inhibits their calcification potential, whilst the activation of one pathway may result in the undesirable inhibition of the activin signaling pathway increases calcification. activation or inhibition of another pathway. Thus, cross-regulatory It is still largely unclear, however, how the underlying signaling interconnections likely have evolved hand-in-hand with insulating pathways lead to the observed biological outcomes. Even mechanisms that function to limit undesirable spillover. Our though many experimental and theoretical studies have revealed objective is to understand the mechanisms that have evolved to detailed mechanistic characteristics of signal transduction maintain specificity from signal to cellular response, and to limit networks in general, little is known about the mechanisms that ‘leaking’ between pathways, despite extensive interconnections leads to the actual specific cellular response. Using a combined and component sharing. theoretical and experimental systems biology approach, we Results: Our experimental work has focused on the mitogen- are aiming to bridge this gap by constructing a quantitative activated protein kinase (MAPK) cascade regulating mating, model of the dynamic behaviour of the signaling events in a stress responses and invasive growth in budding yeast. We have well-defined osteoblast differentiation system and relate this to delineated several specificity promoting mechanisms, including the mineralization process, with read-outs both at the level of (1) docking interactions between MAPKs and their substrates and transcriptome and proteome responses, in order to bring to light regulators; (2) feedback circuits that control the magnitude and key differences capable of controlling cell fate. duration of pathway activation; (3) selective activation of scaffold proteins. At the same time, we have developed a theoretical DS2-1-74 framework for the analysis of specificity in cell signaling, and have developed mathematical models of several important insulating Identification of key nodes in signal transduction network mechanisms. of cell differentiation Conclusions: We conclude that networks in which different Kel, Alexander pathways share components will likely contain one or more BIOBASE GmbH, R&D, Wolfenbuettel, Germany insulating mechanisms that enhance specificity and fidelity. Furthermore, more than one insulating mechanism will probably Objective: Current studies of the molecular mechanisms of be required to provide a reasonable degree of mutual specificity complex cellular processes such as cell differentiation boil down and fidelity without placing significant constraints on the network. to analysis of complex networks of protein-protein and protein- In the yeast mating/invasive growth network, the insulating DNA interactions which combines signal transduction and mechanisms of cross-pathway inhibition and selective scaffold transcription factor networks. Searching for key nodes in such activation work together to provide signaling specificity and network will help us to understand the differentiation program. fidelity. Results: We developed a novel computational tool, ExPlain™

110 ICSB 2008 DS2-1-76 Biology Markup Language (SBML) for dynamics simulations. In addition, the map will be published on the online, community Sequence and structure requirements for miRNA- tagging platform Payao and researchers will be able to view and dependent RISC binding and mRNA degradation update the map in an iterative and collaborative manner. Hausser, Jean1; Landthaler, Markus2; Gaidatzis, Dimos3; Tuschl, Conclusions:This crosstalk map provides an integrative view Thomas2; Zavolan, Mihaela3 of the molecular interactions characterizing two key pathways 1Biozentrum, Universität Basel and Swiss Institute of involved in cancer pathogenesis. Structural analysis as well Bioinformatics, Basel, Switzerland; 2Howard Hughes Medical dynamical simulations based on the map can provide insights Institute, Laboratory for RNA Biology, The Rockefeller University, into potential molecular mechanisms contributing to robustness New York, United States; 3Biozentrum, Universität Basel, Basel, of cancer cells and can potentially help to search for new drug Switzerland targets against cancers resistant to endocrine therapy.

Objective: MiRNAs are important regulators of mRNA translation DS2-1-78 and stability in plants and animals. Initial studies indicated that miRNAs induce translational repression, while more recent Towards understanding global regulation of drug experiments demonstrated that miRNAs also trigger degradation detoxification in primary human hepatocytes - Inference of their mRNA targets. Combining an Ago-Immunoprecipitation from perturbed expression profiles (IP) approach for miRNA target identification with computational Murugan, Prem Kumar1; Reichart, Thomas2; Bonin, Michael3; analysis, we investigate what sequence and structure features of Reuss, Matthias1 the miRNA targets determine the efficiency of RISC binding and 1Institute for Biochemical Engg., University of Stuttgart, Stuttgart, the subsequent degradation of the mRNA. Germany; 2Institute for Technical Biochemistry, Universtiy of Results: We demonstrate that the accessibility of the miRNA Stuttgart, Stuttgart, Germany; 3Microarray Facility, Medical target site and the energy of miRNA “seed” interaction with Genetics, University of Tuebingen, Tuebingen, Germany the target influence the efficiency of RISC binding, whereas the sequence composition of the environment of the miRNA target Objective: The fundamental objective of this study was to obtain site influences the rate of subsequent mRNA degradation. a larger picture of the key components involved in the regulation We further find that compositional biases that promote mRNA of CYP3A4 in primary human hepatocytes upon induction with degradation also characterize the environment of predicted the clinical drugs, Rifampicin, and Statins. Due to a vast number miRNA target sites that are evolutionarily conserved, suggesting of unknowns, a top down approach was decided to be the that the miRNA-induced degradation of target mRNAs is a best possible solution to study this phenomenon under a global common phenomenon that has been selected evolutionarily. context. The motivation to use coarse methods of network Posters Combining a small subset of the features that we studied, we can reconstruction under the Boolean framework was to underpin Dedicated train a model to predict degradation of mRNAs carrying miRNA the interactive nature of a multitude of components from the target sites at an accuracy comparable to that of a program standpoint of multivariate relationships rather than the classical that is most used currently. Furthermore, we are able to predict univariate analysis. This facilitated the study of the autonomy of miRNA targets sites that have only limited complementarity with subnetworks that were grown from the inferred interactions, since the miRNA 5’ ends. time and again, it has been well established that a small set of Conclusion: We show that sequence and structure features genes maintain the core regulatory mechanism. of miRNA target sites and of their immediate neighborhood Results: Reconstruction of the CYP3A4 regulatory network contribute to different extents to individual steps in the miRNA- shows new interactions that were hitherto unknown. This is due dependent silencing cascade. Furthermore, we show that to the fact that the combinatorial aspect of the approach used evolution favored sites whose sequence environments are in this study was able to identify them compared to the classical supportive of mRNA degradation. These environments are mostly single gene correspondence. Likewise, interaction relationships U-rich and are probably bound by various RNA-binding proteins were constructed for the other three targets, PXR, MDR1, and that may modulate miRNA activity. RXR. It was also noted that some interactions, though of lesser magnitude, in the order of 25-30%, were still extremely crucial DS2-1-77 to preserve the autonomy of the network. This was reflected on the strength of connection between PXR and HNF4a in their A comprehensive molecular map of crosstalk between the interaction with CYP3A4. PGC1, SHP, and SRC1, the newly EGFR and the ER pathways in cancer cell lines adjoined genes in the subnetwork had high interacting strengths Suzuki, Ken; Ghosh, Samik; Matsuoka, Yukiko; Kitano, Hiroaki with both PXR and HNF4a, making them an integral part of The Systems Biology Institute, Tokyo, Japan the autonomous regulatory circuit. A similar analysis was also conducted with Statins being the inducers. Objective: Resistance to anticancer therapy is one of the Conclusions: The aforementioned methods provided insights most important topics in cancer biology and clinical medicine. into the topology, and the interaction strengths of the gene Among many types of cancers, breast cancer is a leading regulatory network governing drug metabolism. Subnetworks cause of cancer death, and its resistance to endocrine therapy were especially interesting to investigate network autonomy and is an unsolved problem. Clinical and experimental evidence pinpoint the roles of newly identified key partners in such an suggests that the cross-talk between the EGFR and the ER autonomous network. pathway plays a key role in this phenomenon. While researchers have accumulated much knowledge, the whole image of the DS2-1-79 network and the mechanism causing drug-resistance is not fully understood. Thus, building a holistic map of the interaction The role of G-proteins mediating the polarization response between EGFR and ER pathways is a key step in understanding of yeast cells to changing α-factor gradients cancer pathogenesis. Moore, Travis1; Chou, Ching-Shan2; Jeon, Noo Li3; Nie, Qing2; Yi, Results: In this work, we present a comprehensive molecular Tau-Mu1 map of crosstalk between the epidermal growth factor receptor 1University of California, Irvine, Developmental and Cell Biology, (EGFR) and the estrogen receptor (ER) signaling network for Irvine, United States; 2University of California, Irvine, Mathematics, cancer cell lines. Based on previously published molecular map Irvine, United States; 3University of California, Irvine, Biomedical of the EGFR signaling system (Oda. et al), we constructed a Engineering, Irvine, United States map of the ER pathway extracting information from literature on MCF-7 breast cancer cells. The map identifies the structure In changing environments, organisms orient themselves to of the interaction between the two cascades. It is created in dynamic external cues. It is necessary to sense and respond to CellDesigner software (Version 4.0 beta) using a well-defined and the environmental signal and also to track changes in the signal. consistent graphical notation and can be stored in the Systems During mating of the budding yeast S. cerevisiae, haploid cells

ICSB 2008 111 sense the presence and direction of cells of the opposite mating We demonstrate that given some of those mechanisms the type via spatial gradients of mating peptide, and then mount a timescales observed in ChIP can be reconciled with those response by projecting toward their potential mate. For the two known from photobleaching experiments. A stochastic model cells to sense the other’s direction, project and meet precisely, was constructed that shows that the observed co-factor cycling tip to tip, they must undergo constant environmental sensing can arise in a cell population due to the sequential nature of and shmoo steering. To examine the ability of yeast to actively transcription initiation. We also demonstrate that the PDK4 mRNA sense their environment, we exposed cells to microfluidically- accumulation observed experimentally is consistent with the generated spatial gradients of mating pheromone and then model predictions. switched the gradient direction 180o. We observed that cells Conclusions: Transcription initiation on the PDK4 gene promoter exposed to low concentrations of mating factor continued to was experimentally shown to be a cyclic process which is extend their projection, which bent in response to the directional consistent with findings in other systems. In addition, it was switch, whereas cells exposed to high concentrations of mating demonstrated that corresponding cycles can be also found at the factor formed a second projection in the appropriate direction. mRNA level. To our knowledge this is the first mechanistic model Both heterotrimeric and Cdc42 G-protein systems are required explaining the observation of such cycles. for sensing and responding to spatial gradients, and we studied the response of mutants within these two pathways as well as at DS2-1-82 their interface. Computational modeling techniques were used to explore the balance between these systems and how they Structural analysis of toll-like receptor signaling network mediated the ability of cells to sense and initially polarize in a Ozbabacan, Ece; Tekir, Saliha; Ulgen, Kutlu O. robust way, and yet retain the ability to track directional changes Bogazici University, Chemical Engineering, Istanbul, Turkey in the external signal as they occur. Objective: Comprehensive maps of large-scale signaling DS2-1-80 pathways are attractive for researchers interested in signal transduction mechanisms because the information they give can How quantitative measures unravel design principles in be used as a basis for mathematical interpretations about the multi-stage phosphorylation cascades topological structure. The objective in this study is to analyze Dedicated Frey, Simone1; Millat, Thomas1; Hohmann, Stefan2; Wolkenhauer, the comprehensive TLR signaling map structurally, in order to Posters Olaf1 enlighten its topological structure, i.e. the properties such as 1University of Rostock, Rostock, Germany; 2Göteborg University, robustness, and the detection of the crosstalking molecules in Göteborg, Sweden sub-pathways of the large system. Ultimate goal is then to identify these key molecules as potential drug targets for the infectious We investigate design principles of linear multi-stage diseases. phosphorylation cascades by using quantitative measures Results: Network topology is investigated using the analyses for signaling time, signal duration, and signal amplitude. We based on graph theory and linear pathways (Watts and Strogatz, compare alternative pathway structures by varying the number 1998; Albert et al., 1999; Barabási, 2002). The TLR network is of phosphorylations and the length of the cascade. We show found to be scale-free and it has small-world properties. Small- that a model for a weakly activated pathway does not reflect the world networks are described as having many molecules with biological context well, unless it is restricted to certain parameter few interactions, but few highly connected molecules (hubs). combinations. Focusing therefore on a more general model, According to the analysis based on graph theory, in- and out- we compare alternative structures with respect to a multivariate degree distributions tend to fit the power law model (P(k)≈k-γ) optimization criterion. We test the hypothesis that the structure with γvalues of 2.12 and 2.14 and R2 values of 0.96 and 0.92, of a linear multi-stage phosphorylation cascade is the result of an respectively. The hubs for the in-degree connections are TLR2 optimization process aiming for a fast response, defined by the ligand and IKK-α/IKK-β/IKK-γ complex (which is also a hub minimum of the product of signaling time and signal duration. It is for the out-degree connection), whereas the hubs for the out- then shown that certain pathway structures minimize this criterion. degree connections are found as TIR domain, TRAF6, Rac1/ Several popular models of MAPK cascades form the basis of our GTP complex, κB site/NF-κB (p65)/NF-κB (p50)/CBP complex, study. These models represent different levels of approximation, and MyD88. The network diameter of TLR signaling system is which we compare and discuss with respect to the quantitative calculated as 25 with a mean path length of 8.98. measures. Conclusions: Due to the small-world topology of the network, it can be concluded that the TLR network is highly robust, i.e. it can DS2-1-81 take alternative actions under perturbations. The critical signaling molecules, i.e. common species participating in most of the The mechanism of nuclear receptor PPARβ/δ mediated pathways leading to the phenotypes, are found to be: Rac1/GTP transcription activation of the human PDK4 gene complex, TLR2/TLR2 ligand complex, and Ubc13/Uev1A/TRAF6/ Rybakova, Katja N1; Degenhardt, Tatjana2; Mone, Martijn J.1; TAB2/TAB3/TAK1/TAB1 complex. Bruggeman, Frank J.1; Westerhoff, Hans V.1; Carlberg, Carsten2 1Vrije Universiteit, Amsterdam, Netherlands; 2University of Kuopio, DS2-1-84 Kuopio, Finland Modeling, analysis and control of the translational initiation Objective: Eukaryotic transcription regulation is an intricate regulation process in eukaryotic cells process that requires coordinated action of many proteins, Bar, Nadav S. involved in chromatin re-modelling and pre-initiation complex Norwegian University of Science and Technology, Department of formation. Many mechanisms, especially of initiation induced by Chemical Engineering, Trondheim, Norway distant regulatory elements, remain elusive. The human PDK4 (pyruvate dehydrogenase kinase 4) gene presents a convenient Objective: The main objective is to model and control the model system to study the initiation mechanisms as it relies on a process of initiation regulation of protein synthesis in eukaryotic single distant response element (RE) controlled by PPARβ/δ. We cells. It was found in the last decade that gene regulation use this experimental system to study the temporal aspects of in the level of translation is mainly effected by the initiation transcription initiation. process. Eukaryotic initiation factors (eIF-) 2 and -4 are the main Results: ChIP assays used to assess the identity and abundance components that regulate gene expression, by enabling the 40S of regulatory proteins demonstrated cyclical protein binding ribosomal subunit structure to bind to a pre-initiation complex eIF- on both the transcription start site (TSS) and on the RE during 4F. This process is complex and involves many different regulating transcription initiation. These results were further analyzed components. Tools such as a dynamic model, analysis and with kinetic models. We postulate alternative mechanisms for simulations help to understand this complex process. pre-initiation complex assembly and test them for feasibility. Results: This paper presents ordinary differential equations (ODE)

112 ICSB 2008 models for the eIF-2 and -4 regulation processes and analysis of activated EpoR generates a feed-forward loop to amplify the these . A further reduction of the eIF-2 process is presented and activation of Akt. nonlinear analysis proves L2 stability of the process. Simulations For Akt activation, the data-based mathematical model revealed and analysis of the reduced model led to several conclusions that the time course data can only be appropriately captured about the affect of a few principle factors in the eIF-2 initiation if negative regulation triggered by the EpoR is assumed. process (namely eIF-2B, phosphorylated and non-phosphorylated Immunoblotting data from murine erythroid progenitor cells at eIF-2 and GCN2), consistent with recent experiments in yeasts. A the colony forming unit-erythroid stage (CFU-E) and BaF3 cell nonlinear feedback control loop using measurements from three lines revealed that the different endogenous expression levels states was proposed, which can drive the production of protein of the phosphatases SHIP1 (SH2 domain containing inositol- to a desired set-point. However, despite the fact that It is possible 5-phosphatase) and PTEN (phosphatase and tensin homolog) to measure these three factors using recent technology, frequent coincide with a different time delay of down-regulation of measurements is needed (seconds) to satisfy the control law. Akt phosphorylation. Furthermore, by overexpression of wild Conclusions: It is theoretically possible to control the process type SHIP1 or PTEN in CFU-E and BaF3 cells we observed of protein synthesis in eukaryotic cells by manipulating the intake a dramatic reduction of Akt phosphorylation. On the basis of of eIF-2 factor, driving the protein production to a desired value. these experimental findings and the corresponding extension of The main limitation of this approach is the required time series our mathematical models, we identifed the negative regulatory measurements. mechanism acting on the PI3K pathway. Conclusions: Based on our mathematical models, we propose DS2-1-85 that for early phase of signalling activation, the activated receptor generates a feed-forward loop to amplify the activation of Akt and Type I IFN feedback-dependent and independent antiviral at later point, inhibitors of signaling, are activated, such as the gene program in innate immune cells lipid phosphatases PTEN and SHIP1. Kumagai, Yutaro; Takeuchi, Osamu; Akira, Shizuo Immunology Frontier Research Center, Osaka University, DS2-1-87 Laboratory of Host Defense, Suita, Japan Elucidating the HPA axis stress response via computational Innate antiviral immune system invokes pleiotropic responses inverse analysis including production of an important antiviral cytokine type I Zarzer, Clemens1; Machne, Rainer2; Koehler, Gottfried3; Lu, interferon (IFN) through various signaling pathways involving James1 innate immune receptors. Innate immune receptors recognize 1Johann Radon Institute for Computational and Applied 2 viral components induce type I IFN. The induced type I IFN Mathematics, Vienna, Austria; University of Vienna, Theoretical Posters elicits not only antiviral gene induction but also is said to drive Biochemistry Group, Vienna, Austria; 3University of Vienna, Max F. Dedicated positive feedback loop. However, the contribution of feedback Perutz Laboratories, Vienna, Austria signaling and innate receptor signaling in induction of antiviral gene expression is still obscure, especially in vivo situation. By Objective: he hypothalamic pituitary adrenal (HPA) axis taking advantage of comprehensive gene expression profiling in represents a feedback system that plays an important role combination with newly generated technology that enables single in maintaining the body homeostasis in response to various cell-level monitoring of IFN-α induction in vivo, we identified a set stresses. When stress is encountered, the hypothalamus of genes co-regulated with type I IFN. Interestingly, these type I releases the corticotropin releasing hormone (CRH) as a central IFN co-regulated genes formed a cluster clearly distinguishable neuro-transmitter in the HPA axis. There exist diverse differential from a cluster of genes whose induction was solely dependent equation models, which account for induction of ACTH synthesis on type I IFN receptor signaling. These type I IFN-inducible genes in the pituitary by CRH, leading to adrenal activation and were upregulated even in cells that did not produce type I IFN, release of cortisol, which in turn inhibits the synthesis of ACTH. indicating type I IFN activated bystander cells. In vivo monitoring Starting from these basic models, several additional feedback of IFN-α production further showed that type I IFN feedback mechanisms could be included. One is the incorporation of an was required on inactivated viral infection or synthetic ligand additional membrane bound glucocorticoid receptor (GR) in the stimulation but not on live virus infection. These results together inhibition of the ACTH release in the pituitary, responsible for fast indicated that innate receptors directly invoke antiviral gene feedback effects. Including such model extensions, computational program, and type I IFN receptor signaling activates bystander inverse analysis is crucial in identifying the possible dynamical cells as well as conditionally drives positive feedback loop. behaviors, such as oscillations,modulated by circadian rhythms and switching between multiple steady states. DS2-1-86 Results: To identify factors controlling the qualitative nature of the stress responses, we apply the method of inverse bifurcation Dynamic modeling of feedback regulation of PI3K signaling analysis, using a hierarchical identification strategy based in the hematopoietic system upon a sparse-promoting regularization method. In particular, She, Bin1; Wang, Lu2; Schilling, Marcel1; Salazar, Carlos2; Hoefer, diseased phenotypes as represented mathematically by the Thomas2; Klingmueller, Ursula1 respective bifurcation diagrams are computationally mapped to 1German cancer research center, Systems biology of signal the underlying regulation mechanisms. For instance, the identified transduction, Heidelberg, Germany; 2German cancer research mechanisms underlying the delayed activation of the stress center, Modeling of Biological Systems, Heidelberg, Germany response include the degradation rate of GR as well as the rate of up-regulation in the GR synthesis via its dimer. Objective: In the hematopoietic system, ligand binding to the Conclusions: The HPA axis involves complex nonlinear feedback Epo receptor (EpoR) activates the phosphoinositide kinase 3 reactions, whose behavior is difficult to understand via time- (PI3K) pathway, which is of special importance in the transduction course simulations alone. By using inverse bifurcation analysis, of proliferative signals. In our study, quantitative immunoblotting we identify important mechanisms governing the qualitative was used to measure activation kinetics of pathway components, dynamics of the HPA axis response to stress as well as propose and mathematical models for the PI3K pathway were developed mechanistic details of the model to be elucidated via experimental in parallel. studies. Results: Signaling through PI3K can be activated either by direct binding to the phosphorylated EpoR at position Y479 or indirectly by the adaptor protein Gab1/2. To determine the contribution of the two mechanisms of PI3K signaling, we expressed different EpoR mutants in the lymphoid pro-B cell lines (BaF3) to selectively activate one of the pathways. The mathematical model established based on these experimental data indicates that the

ICSB 2008 113 DS2-1-88 behavior of a SC or TAC: A cell cycle length and a probability, p, that the progeny of a division remain of the same cell type, Quantitative modeling of synchronous oscillations in delta- versus moving to the next lineage stage. Although feedback notch signaling that increases cell cycle length can improve response times, we Theis, Fabian1; Tiedemann, Hendrik2; Schneltzer, Elida2; find that negative feedback ontop produces strikingly useful Przemeck, Gerhard2; Hrabé de Angelis, Martin2 behaviors. These include a robust steady state, rapid response 1Helmholtz Zentrum München, Institute of Bioinformatics and times, and a lack of requirement for SCs to be programmed to Systems Biology, Neuherberg, Germany; 2Helmholtz Zentrum self-replicate precisely half the time. Under some circumstances, München, Institute for Experimental Genetics, Neuherberg, sustained long-period oscillations can be produced, much like Germany those observed in hair follicle growth and blood cell renewal. In the OE, modeling suggests that interlocking feedback loops Objective: Oscillatory networks are a key motif in many gene aimed at distinct amplifying cell stages may serve to enable the regulatory systems such as cell cycle, circadian clock and tissue to regenerate as rapidly as possible after perturbations over somitogenesis, the focus of this contribution. How oscillation a wide variety of magnitudes. in this multi-cellular system is robustly maintained over more Conclusions: Combining feedback growth regulation with multi- than 80 hours is a fundamental question in the field. It has been stage TAC lineages creates powerful strategies for controlling shown that Delta-Notch signaling is a key functional element tissue size and growth. in the synchronization of the oscillating Hes-genes. We study an extended version of the dynamical system put forward in DS2-1-90 (Tiedemann et al. J Theor Biol 248(1):2007), which models the oscillator as an ordinary differential equation of Hes1-mRNA and Dynamic simulation of Toll-like Receptor 3 signaling -protein concentrations in the two compartments nucleus and pathway cytoplasm. The question is how to formulate and automatically Helmy, Mohamed; Tomita, Masaru; Tsuchiya, Masa; Selvarajoo, determine parameters to allow for ongoing oscillation and efficient Kumar synchronization via the Delta-Notch pathway. Institute for Advanced Biosciences, Keio University, Yamagata, Results: We propose a new measure for synchrony of the Japan Dedicated system that we then optimize via global optimization methods. Posters The measure consists of three terms that are being weighted Objective: The Toll-like receptor (TLR) 3 plays a critical role accordingly to guarantee approximately equal contribution: in mammalian innate immune response against viral attacks (i) each Hes1-oscillator is to produce minimally decaying by recognizing double-stranded RNA (dsRNA) or its synthetic oscillation with maximal amplitude, (ii) the oscillation frequency analog polyinosinic-polycytidylic acid (poly (I:C)) and leads to the needs to be as close to 120 minutes (for mouse) as possible, induction of type I interferons and proinflammatory cytokines. (iii) synchronization of the coupled system is to be as fast as Although the individual components of TLR3 signaling have possible. Experimental data is used to determine some rates been extensively studied, the collective behavior of this pathway in the system. After minimizing the cost function in terms of remains a major challenge. the other parameters, we achieve synchronization within 2-4 Results: Here we used dynamic computational model developed oscillation periods (frequency 121 minutes) in a two-cell system. by E-Cell, to analyze the TLR3 pathway. The model was When extending this to a line of cells, we observe a similar created based on first-order linear response and experimental synchronization behavior but only after 10-15 oscillations. activation profiles of NF-kappaB and MAP kinases in wildtype, Conclusions: Quantifying synchronous behavior in TRAF6 knock-out (KO), RIP1 KO and TAK1 KO cells to poly somitogenesis in a robust and efficient way allows us to optimize (I:C) stimulation. Our results show three important novel features a coupled cellular system for maximally fast synchrony. The of TLR3 pathways. Firstly, the comparison of static signaling employed global optimization method based on a combination topology with experimental profiles suggests poly (I:C) stimulation of differential evolution and simulated annealing robustly avoids activates TAK1, RIP1 and TRAF6 independent from each other. local minima, at the same time allowing simple specification of Secondly, analyzing the time to reach peak level activation for NF- biologically meaningful parameter boundaries. kappaB and JNK in wildtype cells, the dynamic model suggests uncharacterized novel intermediates operate upstream of TLR3. DS2-1-89 Finally, simulation of the different experimental dynamics of NF- kappaB and JNK activities in TRAF6 KO, RIP1 KO and TAK1 Feedback control of proliferative dynamics in multi-stage KO predicts the existence of a novel pathway from TRIF to MAP cell lineages kinases activation, but not NF-kappaB. Lander, Arthur1; Gokoffski, Kimberly2; Lo, Wing-Cheong3; Chou, Conclusion: Thus, our systemic approach provides better Ching-Shan3; Wan, Frederic3; Nie, Qing3; Calof, Anne2 understanding of sophisticated immune signaling pathways. 1University of California, Irvine, Center for Complex Biological Systems, Irvine, United States; 2University of California, Irvine, DS2-1-91 Department of Anatomy and Neurobiology, Irvine, United States; A cross-inhibitory positive feedback mechanism 3University of California, Irvine, Department of Mathematics, Irvine, establishes the robust sharp border of the forebrain United States choroid plexus Srinivasan, Shyam1; Currle, David2; Hayes, Wayne3; Monuki, Objective: In self-renewing tissues (skin, gut, blood, etc.), cells Edwin4; Lander, Arthur5 are usually generated by lineages in which a stem cell (SC) 1University of California, MCSB, CCBS, Irvine, United States; produces to one or more stages of “transit amplifying” cell (TAC), 2University of California, Developmental and Cell Biology, Irvine, the last of which gives rise to a differentiated cell. Accumulating United States; 3University of California, Computer Science, Irvine, evidence indicates that proliferation in such lineages is controlled United States; 4University of California, Pathology, Developmental by feedback factors made by the cells themselves. In some and Cell Biology, Irvine, United States; 5University of California, tissues, such as the mammalian olfactory epithelium (OE), Developmental and Cell Biology, CCBS, Irvine, United States experimental data on feedback control are sufficiently detailed (see abstract by Gokoffski et al., this meeting) to allow one to Objectives: The forebrain dorsal midline (DM) includes the build reasonable models. We use these models to investigate how choroid plexus (CP), which is marked by the expression of features such as stability, steady-state robustness, and response Transthyretin (Ttr). The CP is crucial for proper brain health time (rate of tissue regeneration) are influenced by system and function. During early development, the DM is a source of architecture (e.g. number of TAC stages), feedback configuration diffusible Bone morphogenetic proteins (Bmps), which participate (which cells produce and which respond to feedback factors), and in patterning the CP and surrounding cortex. We elucidate the parameter values. mechanism that governs the formation of the anterior CP border Results: At least two parameters are required to represent the and investigate its performance objectives.

114 ICSB 2008 Results: We provide experimental evidence of cross-repression DS2-1-93 between Bmps and Fibroblast growth factors (Fgfs) produced at the rostral midine. First, using explant studies, we show that Multi-site differential regulation in nuclear receptor Bmps induce Ttr expression while Fgfs repress it. Recent studies signaling that show Fgfs repressing Smad1, a downstream effector of Bmp Kolodkin, Alexey1; Bruggeman, Frank J.2; Moné, Martijn J.1; signalling, provide a basis for understanding Fgf’s repression Bakker, Barbara B.1; Westerhoff, Hans V.3 of Ttr. Secondly, qRT-PCR studies in our lab indicate that Bmp 1VU University of Amsterdam, Molecular cell Physiology, inhibits Fgf’s target genes. Through modelling, we then show that Amsterdam, Netherlands; 2Netherlands Institute for Systems the Bmp-Fgf cross-repression forms a positive feedback loop (PF) Biology, Multiscale modelling and nonlinear dynamics, that makes and positions the anterior CP border. The PF model is Amsterdam, Netherlands; 3University of Manchester, Manchester supported by our study that shows Msx1 having an ultrasensitive Centre for Integrative Systems Biology, Manchester, United response to Bmp in the presence of Fgf. Finally, we compare the Kingdom PF model to other well studied border formation mechanisms, in silico. We find that PF performs a dual role in making the border Objectives: Nuclear receptors (NRs) are part of a large position more robust to morphogen fluctuations and in making a transcription factor family that regulate gene expression in a sharper border. ligand-dependent fashion. Many NRs, including VDR, PXR and Conclusions: Bmp and Fgf cross repress each other’s signalling the steroid receptors, reside in the cytoplasm prior to activation pathway to form a PF loop that makes border positioning robust and translocate to the nucleus upon ligand addition, hence and sharper borders. Developmental systems are replete with resulting in regulation of gene expression. In many cases (e.g. instances of borders formed by cross-inhibitory PF. Furthermore, PXR, VDR and GR) ligand binding strongly affects both nuclear the Bmp-Fgf interaction occurs in many developmental contexts import and export, resulting in active nucleo-cytoplasmic such as limb, teeth, and retinal development. Thus, our findings shuttling. In addition, ligand-dependent changes to the NRs may be of help in understanding the oft-repeated themes of Bmp- moderate their affinity for response elements on the DNA. Fgf interaction and cross-repression in biological systems. Consequently, NR signaling as a whole appears to be regulated by a multi-site control mechanism, i.e. making the concerted DS2-1-92 action of ligand-dependent regulation of nuclear import and export, as well as of DNA binding, crucial in the regulation of Transcriptional network of Drosophila neurogenesis is gene expression. We analyzed this aspect of signaling using both robustly designed detailed and simple kinetic models and show the benefits of this Nakajima, Akihiko; Ishihara, Shuji regulatory design for network functioning.

University of Tokyo, Department of Basic Science, Tokyo, Japan Results: Our simulations demonstrate that multi-site regulation Posters is responsible for a synergy in regulation: the summed regulatory Dedicated Objective: Neurons in the ventral nervous system of Drosophila effect is larger than the effect of each regulatory mechanism are derived from the neural stem cells called neuroblasts. The in itself. We show its relevance for the particular case of neuroblasts show the expression of the transcription factors glucocorticoid receptors. Cytoplasmic localization of unliganded Hunchback (Hb), Krüppel (Kr), Pdm and Castor (Cas) in a receptor protects cells from undesired activation of gene sequential manner along with cell divisions. Each transcription expression. Furthermore, high shuttling flux make the signaling factor specifies the neuronal fates of the progenitor cells. The sensitive to small ligand-dependent changes in nuclear import temporal expression patterns of mutants as well as wild type and export rates. are well examined, but the underlying transcriptional regulation Conclusions: Multi-site regulation can be advantageous through is not fully understood. In addition, for the reliable formation of synergy and embodies an important design principle in NR the nervous system, the susceptibility of the expression to the signaling. perturbations is important issue. Results: By computational modeling of the neurogenesis DS2-1-94 transcriptional network, we addressed the above issues. First, we exhaustively searched for all of the possible network Novel interaction partners for Ruk/CIN85 SH3 domains topologies that can reproduce the expressions of wild type and identified by LC-MS/MS mutants. None of the possible networks consisting only of the Havrylov, Serhiy; Redowicz, Jolanta transcription factors Hb, Kr, Pdm and Cas, can reproduce the Nencki Institute of Experimental Biology, Department of required expressions. However, with the addition of two unknown Biochemistry, Warsaw, Poland regulators, we found the networks that can perform the required patterns. Next, we analyzed the susceptibility of the expression to Objective: Ruk/CIN85 is a mammalian adaptor/scaffold protein the change of the kinetic constants for these networks. We found with several modular protein interaction domains. These domains that the network topology which agrees with the experimentally endow Ruk/CIN85 with ability to interact with diverse proteins known regulations show correct pattern in wider parameter range and thus determine its promiscuous behavior in several cellular than other network topologies. Furthermore, we identified the processes including receptor endocytosis, PI 3-kinase signalling, several transcriptional regulations that are responsible for the cytoskeletal rearrangements and apoptosis. However, precise stability of the temporal expression pattern. roles of Ruk/CIN85 in these cellular processes remain elusive and Conclusions: We predict the unknown transcriptional regulations identification of full list of interaction partners is required to gain in the neurogenesis network for the explanation of the mutant better insight. expressions. In addition, the results suggest that neuroblasts Results: In our present study utilizing simple approach combining generate the neurons correctly under the fluctuations in the GST pull-down assay and tandem mass spectrometry analysis we environment and individual variations. We conclude that the identified a set of proteins potentially involved in interactions with transcriptional network of the Drosophila neurogenesis has SH3 domains of Ruk/CIN85. Most of newly identified Ruk/CIN85 evolved to perform the robust temporal expression pattern. binding candidates are membrane and actin regulators implicated in membrane trafficking and actin cytoskeleton remodeling. Conclusion: Presented data support the idea that Ruk/CIN85 adaptor/scaffold protein acts in different membrane trafficking processes at the interface of membranes and actin cytoskeleton. Further studies are required to establish which of identified interactions are significant in vivo.

ICSB 2008 115 DS2-1-95 Results: A comprehensive dynamic study of multiple parameters (cells, ROS, markers of oxidative stress, and antioxidants) Mathematical modelling of aquaporin-2 trafficking and disturbed in the investigated subsystems was performed. We regulation revealed that oxidative stress induced enhancement of antioxidant Fröhlich, Martina; Klipp, Edda; Schaber, Jörg defence (e.g., ceruloplasmin increase) in blood, but not in ascites. Max Planck Institute for Molecular Genetics, Otto-Warburg- It seems, that ascites environment possess only non-specific Laboratory, Berlin, Germany antioxidant defence. The compartmental model was developed, and its key nodes were also detected. Objective: Aquaporins are membrane proteins which are selectively permeated by water. In our research we will focus DS2-1-97 on Aquaporin-2 (AQP2) which is expressed in collecting duct cells of the human kidney. AQP2 plays an important role in Calcium microdomains at presynaptic active zones of the reabsorption of water which is regulated by the hormone vertebrate hair cells unmasked by stochastic deconvolution Vasopressin. After Vasopressin stimulation a signalling cascade is Bortolozzi, Mario1; Lelli, Andrea2; Mammano, Fabio1 activated within the collecting duct cells and AQP2 is translocated 1Venetian Institute of Molecular Medicine (VIMM), Padova, Italy, to the apical membrane. This results in an increase in the water Padova, Italy; 2University of Virginia Medical School, Department flow through the cells and therefore in the reabsorption of water of Neuroscience, Charlottesville, VA, United States from the urine to the blood. The aim of our work is, to build up a mathematical model, which can be used to gain new insights Signal transduction by auditory and vestibular hair cells involves into the processes of AQP2 regulation and to find potential drug an impressive ensemble of finely tuned control mechanisms, targets. strictly depending on the local intracellular Ca2+ concentration Results: ([Ca2+]i). The study of Ca2+ dynamics in hair cells (and many We will present a mathematical model of the trafficking and other cell types) typically combines Ca2+-sensitive fluorescent regulation of AQP2. We used known information of the involved indicators (dyes), patch clamp and optical microscopy to proteins as well as time series data available in the literature to produce images of the patterns of fluorescence of a Ca2+ build up a simple model of the AQP2 trafficking. The chemical indicator following various stimulation protocols. Here we Dedicated reactions are given as Ordinary Differential Equations (ODEs) describe a novel method that combines fluorescence imaging Posters based on Mass-Action Kinetics. A thermodynamical approach and numerical simulations to effectively deconvolve Ca2+ signals was used to calculate the fluxes through the cell membranes and within cytoplasmic microdomains that would otherwise remain the change in intracellular osmotic pressure. inaccessible to direct observation. The method relies on the Conclusions: We will show that our model can reproduce known comparison of experimental data with virtual signals derived from data. We will demonstrate how this approach can be used to a Monte Carlo reaction—diffusion model based on a realistic gain new insights into the processes of AQP2 trafficking and the reconstruction of the relevant cell boundaries in three dimensions regulation of water homeostasis. by the Matlab software package. The model comprises Ca2+ entry at individual presynaptic active zones followed by diffusion, DS2-1-96 buffering, extrusion and release of Ca2+. Our results indicates that changes of the hair cell [Ca2+]i during synaptic transmission Disbalance between innate immunity response and are primarily controlled by the endogenous buffers both at short antioxidant defence in blood and ascites: Integration of (< 1 micron) and at long (tens of microns) distances from the experimental and mathematical modeling active zones. We additionally provide quantitative estimates of Shatalin, Yuriy1; Naumov, Andrey1; Sukhomlin, Tatyana1; Ermakov, concentration and kinetics of the hair cell endogenous Ca2+ Genadi1; Potselueva, Margarita1; Sharipov, Ruslan2; Yevshin, Ivan2; buffers and Ca2+-ATPases. In our opinion, this approach is of Kolpakov, Fedor2 potentially general interest as it can be easily adapted to the study 1Institute of Theoretical and Experimental Biophysics RAS, of Ca2+-related phenomena in diverse biological systems. Pushchino, Russian Federation; 2Institute of Systems Biology, Novosibirsk, Russian Federation DS2-1-98

Motivation and Aim: Tumour cells proliferation is known to Membrane identity and GTPase cascades regulated by activate innate immunity thus resulting in oxidative stress and toggle and cut-out switches inflammation, which affect many physiological functions. Recently Del Conte-Zerial, Perla1; Brusch, Lutz1; Rink, Jochen2; Collinet, we have observed a reverse dynamics of ROS generating Claudio2; Kalaidzidis, Yannis2; Zerial, Marino2; Deutsch, Andreas1 capacity of phagocytes in blood and ascites of rats during 1Dresden University of Technology, Center for High Performance transplanted Zajdela hepatoma growth. We hypothesised that the Computing, Dresden, Germany; 2Max Planck Institute of reverse dynamics of some physiological parameters may reflect Molecular Cell Biology and Genetics, Dresden, Germany a disbalance in homeostasis of the two physiological fluids. If so, specific non-invasive treatment able to correct the disbalance Objective: Key cellular functions and developmental processes when applied specifically and at appropriate time could support rely on cascades of GTPases. GTPases of the Rab-family the innate immunity efficacy and prevent tissue injury. The main provide a molecular ID-code to the generation, maintenance goal of this work was to check the hypothesis both experimentally and transport of intracellular compartments. Here, we address and theoretically, to detect key nodes of the subsystems (ascites, the molecular design principles of endocytosis by focusing on blood, innate immunity). the conversion of early endosomes into late endosomes, which Methods and Algorithms: Experimental model: Zajdela entails replacement of Rab5 by Rab7. hepatoma transplanted into peritoneal cavity of Wistar rats. Results: We model this process as a cascade of functional Experiment design: simultaneous measurement of multiple modules of interacting Rab GTPases. We demonstrate that biochemical parameters of blood and ascites. Recently inter-module interactions share similarities with the toggle switch elaborated biochemical, spectral and immune assays methods described for the cell cycle. However, Rab5-to-Rab7 conversion were applied to follow concentrations of individual antioxidants, is rather based on a newly characterised “cut-out switch” ROS, metabolites and regulatory protein in dynamics. BioUML analogous to an electrical safety-breaker. Both designs require (http://www.biouml.org) workbench was applied for the formal cooperativity of auto-activation loops when coupled to a large description, compartmental modelling and simulation of pool of cytoplasmic proteins. Live cell imaging and endosome investigated subsystems. The values of modelling parameters tracking provide experimental support to the cut-out switch in were either obtained from literature or calculated from obtained cargo progression and conversion of endosome identity along the experimental data. BeanExplorer Enterprise Edition (http://www. degradative pathway. beanexplorer.com) technology was used for web access to the Conclusions: We propose that, by reconciling module data placed in the BMOND database (http://bmond.biouml.org). performance with progression of activity, the cut-out switch

116 ICSB 2008 design could underlie the integration of modules in regulatory to generate data on the spatial dynamics of signaling molecules cascades from a broad range of biological processes. labelled with green fluorescent protein (GFP) over time and determine the import and export rates under various conditions. DS2-1-99 For reliable parameter estimation it was crucial to use the identical model system and treatments for all experimental approaches. Data-based analysis of network structures for regulation The expression of STAT5-GFP required for live cell imaging and adaptation of the gp130-JAK1-STAT3 signaling was controlled by a tightly regulatable Tet-on system. At low pathway in primary hepatocytes expression levels, this avoids artefacts caused by changes in Bohl, Sebastian1; Theis, Fabian J.2; Busch, Hauke3; Eils, Roland3; the stoichiometry of signaling molecules. At higher expression Timmer, Jens4; Klingmueller, Ursula1 levels of STAT5 we observed severe alterations of the signaling 1DKFZ, Division Systems Biology of Signal Transduction, dynamics, most prominently prolonged activation of the pathway. Heidelberg, Germany; 2Helmholtz Center Munich, CMB, Conclusions: We investigated these effects of STAT5 IBIS, Munich, Germany; 3DKFZ, iBioS, Heidelberg, Germany; overexpression by quantitative immunoblotting and live cell 4University of Freiburg, FDM, Freiburg, Germany imaging combined with data-based mathematical modeling to identify the underlying mechanisms. Since the JAK-STAT pathway Objective: In the early phase of hepatocyte regeneration, is constitutively activated in a wide array of human malignancies interleukin (IL)-6 rapidly activates the gp130-JAK1-STAT3 this approach will help to identify the most sensitive steps of the signaling pathway enabling the previously quiescent hepatocytes JAK2-STAT5 pathway and offer insight into potential novel drug to enter cell cycle progression, a process known as priming. To targets. understand how the priming phase is controlled, we monitor the signaling kinetics and analyze regulating principles capable DS2-1-102 of capturing the measured dynamics of pathway activation by mathematical modeling. Multi-layer differential activation of biological processes in Results: Stimulating primary hepatocytes with different LPS-stimulated macrophages concentrations and varying durations of IL-6, we found a Piras, Vincent1; Tsuchiya, Masa1; Choi, Sangdun2; Akira, Shizuo3; highly adaptive behavior resulting in a transient activation of Tomita, Masaru1; Giuliani, Alessandro4; Selvarajoo, Kumar1 gp130 with a constant time-point of maximal activation. While 1Keio University, Institute for Advanced Biosciences, Tsuruoka, different IL-6 concentrations are maintained in the amplitude Japan; 2Ajou University, Department of Molecular Science and of receptor activation, differences in the activation profile of Technology, Suwon, Republic of Korea; 3Osaka University, the IL-6 receptor are reduced at downstream components. Research Institute for Microbial Diseases, Osaka, Japan; 4Istituto

This causes a robust and transient activation of STAT3 in the Superiore di Sanita’, Environment and Health Department, Rome, Posters nucleus independent of the applied IL-6 concentration. The Italy Dedicated effect of this switch-like activation of STAT3 on gene expression was analyzed by microarrays, which revealed a strong re- Objective: Toll-like receptors (TLRs) sense invading pathogens organization of the expression profile after IL-6 stimulation. Since and trigger innate immune responses. TLR4 activates Myeloid several mechanisms are known to mediate adaptation, we used Differentiation factor 88 (MyD88)- and TIR-domain-containing mathematical analysis to identify the underlying design principles adapter-inducing interferon-β (TRIF)- dependent pathways upon of the signaling network. Based on our findings, we analyzed recognition of lipopolysaccharide (LPS). Recent studies indicate desensitation, sequestration, degradation, and non-competitive immune response is largely pleiotropic involving a myriad of inhibition as regulatory principles of negative feedback loops cellular processes. In order to investigate this, we performed of the pathway, i.e. SOCS3 and SHP2, to meet the measured multidimensional statistical analysis on temporal whole-genome kinetics. expressions (22690 ORFs) of LPS response to 4 genotypes; Conclusions: We show that positive cooperativity between wildtype, MyD88 knock out (KO), TRIF KO, MyD88 and TRIF activated gp130 and phosphorylated STAT3 causes a switch-like double KO (or DKO) macrophages. activation of STAT3 whereas signal duration is highly dependent Results: Principal component analysis (PCA) and temporal on the expression and half-life of the induced negative feedback Pearson correlation analyses reveal all genotypes’ whole genome regulator SOCS3. expressions possess general similarity especially between MyD88 KO, TRIF KO and DKO. This is surprising as it suggests DKO can DS2-1-101 still respond to LPS in biologically specific manner. Therefore, the presence of MyD88- and TRIF-independent pathways is A systems biology approach to study prolonged activation implicated and further analyses on DKO suggest LPS could of JAK2-STAT5 signaling following overexpression of STAT5 trigger the TLR4/Mal/BTK pathway. Using temporal correlation Pfeifer, Andrea C.1; Posdziech, Florian2; Hengl, Stefan2; Timmer, analysis we also found differential induction of genes in all Jens2; Klingmueller, Ursula1 genotypes occurred early (0-1h) and late (1-4h). These genes can 1German Cancer Research Center, Systems Biology of Signal be hierarchically classified into multi-layer activation of biological Transduction, Heidelberg, Germany; 2University of Freiburg, processes such as wildtype layer, wildtype-DKO layer, etc. (6 Freiburg Center for Data Analysis and Modelling, Freiburg, clusters in total). Finally, similar temporal response occurring Germany between the whole genome and random gene extractions indicates scalable modality of gene expression regulation. This Objective: Signal transduction from the plasma membrane to interesting finding is an emerging biological mode opposed to the the nucleus downstream of many cytokine and growth hormone popular ‘wired’ pathway based approach of cell biology. receptors is mediated by receptor-associated Janus kinases Conclusion: Our system level approach provides several novel (JAKs) that activate transcription factors of the signal transducer insights into the diversity of LPS stimulated cellular response. and activator of transcription (STAT) protein family. By data- based mathematical modeling we previously provided evidence for rapid nucleocytoplasmic cycling of STAT5 with a nuclear residence time of about seven minutes and identified the steps from nuclear import to nuclear export of STAT5 as most sensitive to perturbation within the core module of JAK2-STAT5 signaling. Although the JAK2-STAT5 pathway has been studied extensively in the past, the regulation of these steps is poorly understood. Results: We extended the model of JAK2-STAT5 signaling to include information on nuclear translocation and processing of STAT5. Quantitative time-lapse microscopy and fluorescence recovery after photobleaching (FRAP) in living cells were used

ICSB 2008 117 DS2-1-103 Conclusions: It was confirmed that cells with Snf3 alone respond to glucose at much lower concentrations than do cells with Rgt2 Systematic analysis of the RAS/cAMP/PKA signalling alone. Furthermore, a previously proposed model for transporter- pathway using in vivo fluorescence microscopy like sensors (Wu et al. 2006 J Cell Biol 73:327-331) was tested Bodvard, Kristofer1; Sliwa, Piotr1; Logg, Katarina1; Almquist, by investigating consequences of manipulating intracellular Joachim2; Kvarnstrom, Mats2; Jirstrand, Mats2; Blomberg, sugar concentrations, with results that are consistent with a Anders3; Kall, Mikael1 generalisation of the model to the case of sugars. 1Chalmers University of Technology, Applied Physics, Gothenburg, Sweden; 2Fraunhofer-Chalmers Centre, Gothenburg, Sweden; DS2-1-105 3Gothenburg University, Cell and Molecular Biology, Gothenburg, Sweden Life/Death decision in death receptor-induced apoptosis Lavrik, Inna1; Neumann, Leo2; Pforr, Carina3; Krammer, Peter3; Objective: A systematic approach is used to study dynamics Eils, Roland3 in the RAS/cAMP/PKA pathway in budding yeast by combining 1DKFZ, Bioquant, Heidelberg, Germany; 2Bioquant, Heidelberg, fluorescence microscopy with image analysis and systems Germany; 3DKFZ, Heidelberg, Germany biology. The temporal and spatial information gained from in vivo fluorescence microscopy experiments is excellent for studying Stimulation of CD95 (Fas/APO-1) in some situations results protein expression, dynamics and protein-protein interactions in cell death and in others leads to the activation of NF-κB on a cell to cell basis, which provides input data to build up resulting in cell proliferation. We established an integrated kinetic mathematical models and networks. mathematical model of CD95-mediated life and death signaling. The RAS/cAMP/PKA pathway serves as an important regulator Systematic model reduction resulted in a surprisingly simple of the metabolic and transcriptional activity of yeast cells and is model well approximating experimentally observed dynamics. partially well characterized. However, the regulatory feedback The model postulates a novel link between c-FLIPL cleavage in loop, transferring information from PKA to RAS, remains unclear. the Death-Inducing Signaling Complex (DISC) and the NF-κB In our first approach we are going to perturb the signal flow using pathway. We validated experimentally that CD95 stimulation genetic and chemical methods, and observe resulting changes in resulted in binding of p43-FLIP to the IKK complex followed by its Dedicated GFP-tagged Msn2p localization, which indicates PKA activity. activation. Furthermore, we showed that the apoptotic and NF- B

Posters κ Results: For validation purposes we study the pathway pathways diverge already at the DISC. Model and experimental components and deletion mutations under their native promoters. analysis of DISC formation showed that a subtle balance of We here show how the activation state of the system changes c-FLIPL and procaspase-8 determines life/death decisions in a upon removal of some of the key components, e.g. Δtpk1, non-linear way. This is the first model describing the complex Δbcy1. Following the localization dynamics of Msn2p over time dynamics of CD95-mediated apoptosis and survival signaling. under both different genetic (internal) modifications and under different stress factors, such as light or heat, we detect different DS2-1-106 localization patterns. For instance, no apparent oscillation of Msn2p between the nuclei and the cytosol can be detected for Comparison of two signaling branches of the HOG pathway Δbcy1, a phenomenon that is present for other deletion mutants. by mathematical modeling Conclusions: Introduced signal disruptions can lead to Parmar, Jignesh; Venkatesh, KV; Bhartiya, S clarification of potential interactions between pathway elements. IIT Bombay, Mumbai, India Furthermore, possible physical interactions between proteins are planned to be confirmed by using FRET-FLIM. Continuous Objective: Hyperosmotic shock triggers a signaling pathway oscillations in a biological system are signs that there are one which is known as High Osmolarity Glycerol (HOG) pathway. or several feedback loops. The current observations indicate There are two upstream branches Sln1 and Sho1 which transmit that oscillations in Msn2p localization are driven by one or more a signal to a common MAPKK Pbs2 upon osmotic stress. The feedback mechanisms in the RAS/cAMP/PKA pathway, since utility of multiple branches converging on a single MAP kinase modifications in probable feedback components result in changed and sensing different strength of stimuli is not established, but it Msn2p localization patterns. may help cell to respond over a wide range of osmolarity (Maeda et al. 1995). Although several models have been employed to DS2-1-104 illustrate the mechanism in the HOG pathway, there has not been one that integrates the two branches of the pathway. Our goal is Quantitative characterisation of nutrient sensing by to find out the mechanism of regulation in the signaling pathways transporter-like sugar sensors in Saccharomyces employing multiple branches responding to the same stimuli, by cerevisiae focusing specifically on the osmotic stress response pathway Karhumaa, Kaisa1; Wu, Boqian2; Kielland-Brandt, Morten1 in Saccharomyces Cerevisiae and to extract novel information 1DTU, Center for Microbial Biotechnology, Lyngby, Denmark; on the dynamic operations of the processes underlying osmotic 2Carlsberg Laboratory, Copenhagen, Denmark adaptation. Results: Our model describes the entire HOG pathway Objective: Gene expression in micro-organisms is regulated comprising of two upstream branches, MAPK cascade, and according to conditions such as nutrient availability. Extracellular the downstream effect of the activated Hog1. We study the nutrients can be sensed by integral membrane proteins that are role of two branches Sho1 and Sln1 in osmoadaptation. For present in the plasma membrane. In the yeast Saccharomyces moderate osmotic shock (0.4 to 0.8M NaCl) the difference of cerevisiae, non-transporting sensors with high sequence similarity phosphorylated Hog1 between the Sho1 branch, Sln1 branch to transporters (transporter-like sensors) have been identified for and WT is not much. So for this range of osmotic shock we sugars as well as for amino acids. They initiate a signal of the can say that the two branches are redundant. With increase presence of extracellular nutrients that activates promoters of in osmotic stress, the difference between the amplitude and genes encoding transporters of the respective types of nutrients. duration of phosphorylated Hog1 for Sho1 and Sln1 branches Two transporter-like sensors of sugars, Snf3 and Rgt2, have been increases and the response of Sln1 branch coinciding with that of found in yeast. It has previously been suggested that Snf3 senses WT. This shows that the Sho1 branch is less active under sever lower concentrations of glucose than Rgt2, and that the functions osmotic conditions whereas Sln1 branch functions normally under of the two proteins are overlapping. this circumstances. Results: In this study, a quantitative assay was set up to Conclusion: Our model is consistent with the experimentally directly measure the signalling activity of Snf3 and Rgt2 at proved important features, indicating that it captures some main various extracellular sugar concentrations. Using this assay, we characteristics of the signal transduction along the pathway like: determined apparent affinities of the two individual sensors to (1) Sho1 and Sln1 branch of the HOG pathway are redundant putative elicitors under different growth conditions. for moderate osmotic shock. (2) The Sho1 branch is less active

118 ICSB 2008 under sever osmotic conditions whereas Sln1 branch functions Conclusion: Combining experimental and mathematical systems normally under these circumstances. analyses provided first functional and mechanistic insight into the spatiotemporal coordination of apoptosis execution. Our DS2-1-107 work also provides the first kinetic description of mitochondrial pore formation within living cells and surprisingly shows that this Mathematical modelling of quorum sensing and sporulation key process, which is involved in both caspase dependent and initiation independent cell death pathways, proceeds with kinetics far more Jabbari, Sara1; Heap, John T.2; Minton, Nigel P.2; King, John R.1 heterogeneous than previously assumed. Of note, technically the 1University of Nottingham, School of Mathematical Sciences, approach of rapid imaging and spatial PDE modelling can also Nottingham, United Kingdom; 2University of Nottingham, Centre be applied to many real-time assays targeting fast and complex for Biomolecular Sciences, Nottingham, United Kingdom intracellular signalling processes.

Objective: Sporulation is the process by which symmetrical DS2-1-109 division of a bacterial cell is replaced by asymmetric division, resulting in a mother cell and a forespore. The former goes Modelling Snf1 regulation in saccharomyces cerevisiae on to engulf the forespore, providing it with a coating which is Frey, Simone1; Schmidt, Henning1; Rateitschak, Katja1; resistant to extreme conditions. The decision to enter sporulation Wolkenhauer, Olaf1; Beltran, Gemma2; Garcia-Salcedo, Raul2; is governed by an intricate combination of regulatory genes Elbing, Karin2; Bosch, Daniel2; Ye, Tian2; Hohmann, Stefan2 and proteins which detect the sporulation-inducing signals, 1University of Rostock, Systems Biology and Bioinformatics, one of which is believed to be associated with quorum sensing Rostock, Germany; 2Göteborg University, Department of Cell and (the ability of a cell to detect its population size or density). We Molecular Biology, Göteborg, Sweden mathematically model the corresponding networks in Bacillus subtilis and Clostridium acetobutylicum in order to further our Objective: The AMP-activated protein kinase (AMPK) signalling understanding of the role of quorum sensing in sporulation. pathway plays a central role in monitoring the cellular energy Results: For B. subtilis, the network is well-defined, enabling us status and controlling energy production and consumption. to incorporate four distinct signals: population size (as detected The yeast AMPK orthologue Snf1 is best known for its role in by the putative quorum sensing system, phr/rap), nutrient levels, glucose repression/derepression. One ultimate goal of the study DNA damage and the ability of a cell to become competent. of these pathways is to generate a computational model able The model predicts that the cell will attain high levels of the to support drug development, targeted at advancing diseases sporulation-inducing protein phosphorylated Spo0A only under such as obesity and type-II diabetes. In this work, we focus on appropriate combinations of these signals. For C. acetobutylicum a quantitative dynamic model describing the Snf1 activation/ Posters we focus on the agr quorum sensing system alone and provide deactivation pathway. Dedicated indications that a larger population would be more likely to Results: Experimental data for extracellular glucose, cell density, undergo sporulation than a smaller one. and Snf1-P provided the basis for mathematical modelling. Our Conclusions: The results of our model are consistent with the model is based on ordinary differential equations and correctly hypothesis that quorum sensing systems can serve to activate the reproduces the experimentally measured Snf1 activation and sporulation response in both B. subtilis and C. acetobutylicum. deactivation responses. In order to correctly reproduce in silico Importantly, when B. subtilis has damaged DNA, sporulation will the measured glucose consumption, we had to include synthesis be induced only after a significant time delay, perhaps allowing and degradation reactions for the hexose transporters. the cell time to repair its DNA before undergoing sporulation. Thus Conclusion: The Snf1 signal transduction pathway in yeast the quorum sensing system can act both as a means to detect senses and regulates the energy status of the cell. One aim is to population size and as a timing device. identify yet unknown parts and regulatory loops of the pathway. Key questions concern for instance: Does regulation of Snf1 DS2-1-108 occur via its phosphatase or its kinases or both and how are they regulated? Which role do the hexokinases play? We established Reaction-diffusion analysis of spatiotemporal signaling several model structures that display the different scenarios. The dynamics during ppoptosis execution models were able to reproduce the experimentally measured Snf1 Rehm, Markus; Huber, Heinrich; Hellwig, Christian; Anguissola, activation and deactivation responses. In order to elucidate which Sergio; Dussmann, Heiko; Prehn, Jochen model most likely corresponds to the real regulatory interplay, we Royal College of Surgeons in Ireland, Physiology and Medical suggest further in vivo and in vitro experiments and plan to apply Physics, Dublin, Ireland model discrimination techniques.

Objective: Mitochondrial cytochrome-c release and effector DS2-1-110 caspase activation are rapid processes during apoptosis. Emplying a combination of high-speed cellular imaging Logical modelling of Drosophila mesoderm specification and mathematical modeling we aimed to investigate these Mbodj, Abibatou1; Naldi, Aurélien1; Faure, Adrien1; Junion, fundamental physiological processes at temporal resolutions Guillaume2; Wilczynski, Bartek2; Furlong, Eileen2; Chaouiya, approx. 100 times higher than previously possible. Claudine1; Thieffry, Denis1 Results: Our study provides the first functional and mechanistic 1TAGC - INSERM U928, Marseille, France; 2EMBL, Heidelberg, insight into the spatiotemporal coordination of apoptosis Germany execution: In response to intrinsic or extrinsic stimuli (STS, TRAIL), we were able to mathematically extract the kinetics of pore Objective: The formation of spatially and temporally refined formation in the outer mitochondrial membrane from the initiation patterns of gene expression during embryogenesis is controlled phase of cytochrome-c-GFP redistribution. We also detected by complex regulatory networks. Focusing on early embryonic that the onset of mitochondrial permeabilisation frequently Drosophila development, we have relied on published genetic and proceeds as a wave through the cytosol. Computational analysis functional genomic data (see e.g. Jakobsen et al, 2007. Genes by a partial differential equation model suggested that wave Dev 21: 2448-60) to delineate a logical model encompassing the propagation speeds correspond to diffusion velocities of locally main transcription factors involved in mesoderm specification. generated permeabilisation inducers and may strongly depend on Results: Implemented with the software GINsim (http://gin.univ- their production kinetics. In addition, reaction-diffusion modeling mrs.fr/GINsim/), our logical model accounts for the differentiation predicted that the signaling network of apoptosis execution of mesoderm into four main presumptive territories, each can efficiently translate spatial anisotropies in mitochondrial characterised by a specific combination of active components: permeabilisation into a homogeneous effector caspase somatic muscles, visceral muscles, fat bodies, and heart. Our response throughout the entire cytosol. These model predictions model encompasses many details, making its dynamical analysis corresponded to kinetics we measured in individual cells. challenging. To overcome these difficulties, we have developed

ICSB 2008 119 a method for model reduction, making selected components and their mutual interactions. The model is simulated by means implicit. This approach allows the definition of a full-fledge model of a stochastic algorithm, performed using the tau leaping which can be analysed directly or after reduction. In the reduced procedure. We investigate the system under different conditions, model, interactions are automatically added from the regulators of and we test how different values of several stochastic reaction the implicit nodes onto their targets. The logical rules associated constants affect the pathway behavior. In addition we study the with these targets are revised accordingly. The reduction tool role played by different feedback mechanisms on the behavior of prevents the removal of key components (involved in functional the system. regulatory circuits) to ensure the preservation of the stable states Results: The model is able to simulate in a quantitative way the and most of the dynamical properties. However, subtle delay pathway in a single yeast cell, taking in account also the effect of effects on implicit components can be lost in the reduction molecular noise due to stochastic fluctuations. We were able to process. simulate the Ras protein cycle, the accumulation of cAMP and Conclusions: Simulations with the reduced version of our the activation of PKA, considering a main feedback operating on model qualitatively recapitulates the wild-type situation, as well phosphodiesterase (Pde1) activity. Moreover the addition of a as 27 reported mutant phenotypes (e.g. losses, expansions or Gpr1/Gpa2 module allows to better simulate the cAMP increase transformation of specific territories, following single or combined induced by glucose. The introduction of a feedback on the Ras gene loss-of-functions or ectopic expressions). Furthermore, activator Cdc25 gave only limited effects, instead the introduction in silico experiments with this model tentatively enable the of a feedback on Ira proteins (Ras-GAP) activity changed the delineation of novel experiments. dynamics and under appropriate conditions generates stable oscillations. DS2-1-111 Conclusions: Our model allows a quantitative simulation of Ras/cAMP pathway in yeast and is a tool for testing different Robustness and fragility in the yeast high osmolarity signal hypothesis on the mechanisms that activate and regulate the transduction pathway pathway. Continuous oscillations in the pathway induced by Krantz, Marcus1; Ottosson, Lars-Göran1; Ahmadpour, Doryaneh1; stress conditions has been suggested by experimental data of Warringer, Jonas1; Waltermann, Christian2; Nordlander, Bodil1; Garmendia-Torres et al. (Curr. Biol. 17, 1044, 2007). Blomberg, Anders1; Hohmann, Stefan1; Kitano, Hiroaki3 Dedicated 1Cell- and Molecular Biology, Gothenburg University, Göteborg, DS2-1-113 Posters Sweden; 2Computational Systems Biology, Max Planck Institute for Molecular Genetics, Berlin, Germany; 3The Systems Biology Alternative inputs to α-factor Institute, Tokyo, Japan Tanaka, Hiromasa; Yi, Tau-Mu UCI, Irvine, United States Objective: The cellular signalling networks that integrate various environmental stimuli with information on cellular status must be One of the ultimate goals in systems biology is to understand robust to stimuli fluctuations as well as to stochastic differences signaling networks by experimental and mathematical approaches in the amounts of signalling components. Here, we challenge and to predict outputs of cells from several inputs. The mating the high osmolarity glycerol response (HOG) signal transduction signal transduction system in budding yeast is one of the most pathway in the yeast Saccharomyces cerevisiae with systematic well-analyzed systems, and conventional ways to analyze this disturbances in components’ expression levels implemented by a biological system were mainly done by extracellular inputs “genetic tug-of-war”, or gTOW, methodology. (pheromones or a/α-Factor). Here, we present a novel approach Results: The disturbances were performed under various termed “Alternative Inputs to α-Factor” to dissect and understand external perturbations, including pathway activation by osmotic the mating signaling system. Signaling molecules themselves or shock. The resulting phenotypes in this particular study reflect a constitutively active forms of signaling molecules (from α-Factor wide range of sensitivities, and disperse without any clear pattern receptor (Ste2p) to the transcription factor (Ste12p)) in the over biochemical functions and pathway modules alike, with the mating signaling system were overexpressed in budding yeast, most sensitive nodes being PBS2 and SSK1. and transcriptional activation and the morphological changes Conclusions: Ideally, the obtained sensitivity profiles will allow induced by the alternative inputs were monitored. Interestingly, us to impose parameter constraints. However, a more important these alternative inputs did not induce the same morphology, aspect is the qualitative improvement of model structures, when but induced input-specific morphologies in terms of polarization, local fragilities cannot be explained by the model structure. projection-periodicity, cell size and regulation of cell-cycle arrest. Surprisingly, the “neighboring” nodes HOG1 and SSK2 were For example, AI-Ste12p (Alternative Inputs from Ste12p) and AI- affected to a much lesser extent, questioning our current Ste7p induced the same level of transcriptional activation, but understanding. induced different morphologies from each other. These results suggest that not only signaling strength but also the signaling DS2-1-112 network is responsible for the morphological differences. To investigate the mating signaling system further, each mating Modeling and simulations of the ras signalling pathway in signaling molecule from Ste2p to Ste12p was deleted and each budding yeast: Evidence for oscillatory regimes AI (from Ste2p to Ste12p) was overexpressed in a symmetric Martegani, Enzo1; Pescini, Dario2; Cazzaniga, Paolo2; Mauri, combinatorial manner (“AIs-deletions matrix.”) If this mating Giancarlo2; Besozzi, Daniela2; Colombo, Sonia1 pathway is linear and all AIs are perfect, then the diagonal and 1Università Milano Bicocca, Biotecnologie e Bioscienze, Milano, lower triangle of the AIs-deletions matrix should be 1 (response), Italy; 2Università Milano Bicocca, Informatica Sistemistica e and the upper triangle should be 0 (No response). Transcriptional Comunicazione, Milano, Italy activation almost followed this rule, which suggests that the mating pathway from α-factor to Ste12p is a linear network. Objective: The Ras/cAMP/PKA pathway plays a key role in However, interestingly, mating morphological phenotypes didn’t the control of yeast cell metabolism, stress resistance and follow. These results suggest that the mating signaling system proliferation, in relation to the available nutrients. This signalling possesses greater network complexity than previously thought. cascade is tightly regulated. The experimental studies on the Ras/ cAMP/PKA pathway are complicated by the fact that some key components, like Ras and Cdc25 proteins, are not only required for glucose activation of cAMP synthesis, but they are also essential for basal adenylate cyclase activity and for cell viability. In addition strong negative feedback operates at multiple levels and the experimental perturbations of the pathway often generate contradictory results. Here we use a model that considers the principal components

120 ICSB 2008 DS2-1-114 in epithelial cells. In silico analyses with the software toolbox CellNetAnalyzer revealed phospholipase Cγ1 (PLCγ1) as an Adaptive gene expression to nutrient depletion out of inhibition target to specifically repress theH. pylori induced native regulatory mechanisms activation of the extracellular signal regulated kinase 1/2 (ERK1/2), Tsuru, Saburo; Ying, Bei-Wen; Ushioda, Junya; Kashiwagi, Akiko; whereas the stimulation of ERK1/2 by HGF remains unaffected. Yomo, Tetsuya ERK1/2 activation has been formerly linked to cell scattering in H. Osaka University, Department of Bioinformatic Engineering, Suita, pylori infected cells. The in silico predictions have been confirmed Osaka, Japan experimentally in MDCK cells by using the PLCγ1 inhibitor U73122. Objectives: Living systems have a variety of regulatory Conclusions: This work represents an approach within host- mechanisms for the corresponding surroundings. These pathogen systems biology aiming at deciphering signalling regulatory mechanisms enable cells to adjust the gene expression changes brought about by pathogenic bacteria. The suitability of efficiently when the environment changes. However, it’s unknown our network model is demonstrated by the in silico identification whether cells response to the environmental alterations without of PLCγ1 as a putative target for therapeutic interventions which these inherent regulatory mechanisms, or how cells adjust their is of particular importance in light of emerging resistances against gene expression when encountering the changing environment conventional antibiotics. whose corresponding innate cellular regulatory mechanism or pathway is not yet formed. To discover a possible clue, we DS2-1-116 applied a mini library of bacterial cells holding a broken inherent regulatory path and investigate their expression and growth under Time-resolved description of global phosphoproteome starvation. dynamics for network-wide exploration of novel drug Results: Genes encoding for the enzymes involved in amino acid targets against breast cancer biosynthesis were replaced from the native location in bacterial Oyama, Masaaki1; Kozuka-Hata, Hiroko1; Yumoto, Noriko2; genome respectively to a foreign synthetic gene expression Nagashima, Takeshi2; Kuroki, Yoko2; Inoue, Satoshi3; Hatakeyama, cascade. Cellular expression levels of the featured genes were Mariko2; Kitano, Hiroaki4 quantified in accordance with the fluorescent intensity of the co- 1Institute of Medical Science, University of Tokyo, Tokyo, Japan; expressed GFP. 12 genes corresponding to 5 amino acids were 2RIKEN Advanced Science Institute, Kanagawa, Japan; 3Research examined. We found that some of the featured genes were highly Center for Genomic Medicine, Saitama Medical University, expressed in response to the corresponding amino acid depletion Saitama, Japan; 4The Systems Biology Institute, Tokyo, Japan to recover the cell growth, though the inherent regulatory path was unavailable any more. It indicated that cells were capable Objective: Signal transduction system, in orchestration with Posters to auto-regulate the gene expression level responding to the subsequent transcriptional regulation, widely regulates complex Dedicated changing environments, without the native regulatory systems. biological events such as cell proliferation and differentiation. Subsequently, we analyzed the temporal changes of cellular Therefore, a comprehensive and fine description of their dynamic distribution in gene expression level in the adaptive responses. behavior provides a fundamental platform for systematically We found that the cells proliferated more rapidly as the expression analyzing the regulatory mechanisms that result in each level of the gene increased gradually. Possible models, biological effect. Here we developed a mass spectrometry-based considering stochastic gene expression and the selection process framework for obtaining time-resolved description of global in population level, are proposed to explain such newly achieved phosphoproteome dynamics mediated by serine/threonine/ regulation program. tyrosine residues on cellular signaling molecules. We applied Conclusion: We observed that cells adaptively response our methodology to comprehensive description of the pathways to the surroundings despite of the deficient inherent path. It transmitted through estrogen receptor (ER) or ErbB receptor in suggested that cells could use other paradigms than the known human breast cancer signaling. mechanisms. Possible clues are discussed using stochastic Results: In order to obtain a standard description of breast models. cancer-related signaling networks, we performed time-course analysis of the activation magnitude of phosphorylated proteins/ DS2-1-115 peptides in human MCF-7 cells using highly sensitive nanoLC- MS/MS system in combination with the Stable Isotope Labeling Logical modelling of HGF and helicobacter pylori induced by Amino acids in Cell culture (SILAC) technology. Through c-met signal transduction shotgun identification and quantification of phosphorylated Franke, Raimo1; Gilles, Ernst-Dieter2; Klamt, Steffen2; Poltz, molecules enriched with anti-phosphotyrosine antibodies/ Rainer1; Naumann, Michael1 Phos-tag reagents, we obtained a global view of the dynamics 1Otto von Guericke University, Institute of Experimental Internal of signaling molecules upon estrogen (E2) or heregulin (HRG) Medicine, Magdeburg, Germany; 2Max Planck Institute for stimulation, which clearly revealed the network activation status Dynamics of Complex Technical Systems, Magdeburg, Germany common to or different between these two signaling pathways. Conclusions: Our time-resolved phosphoproteome analysis of Objective: The hepatocyte growth factor (HGF) stimulates human breast cancer MCF-7 cells revealed distinct features of mitogenesis, motogenesis, and morphogenesis by binding the signaling networks driven by the two stimulators. Through to the receptor c-Met. Abnormal c-Met signalling contributes integration with time-course gene expression profiles, our data to tumorigenesis, in particular to the development of invasive should provide an initial basis for constructing a comprehensive and metastatic phenotypes. Abnormal c-Met signalling can be network model to analyze ER/ErbB receptor-mediated signaling induced for instance by the virulence factor cytotoxin associated dynamics at the system-level. We consider that our integrative gene A (CagA) of Helicobacter pylori. This bacterial pathogen has model for breast cancer ER/ErbB signaling will facilitate the colonized the gastric epithelium of half of the human population. analysis of their potential cross-talk machinery which might be Upon translocation into cells, CagA binds to the c-Met receptor involved in the acquisition of drug resistance property against an and promotes cell scattering, which could contribute to the ER antagonist, tamoxifen. invasiveness of tumor cells. We aimed at deciphering the interferences of H. pylori with c-Met signalling and at identifying putative pharmacological targets. Results: We constructed a logical model of HGF and H. pylori induced c-Met signal transduction based on the formalism of Logical Interaction Hypergraphs, which implements the Boolean operators AND, OR and NOT. Relying on quality- controlled published data, the model reveals the differences and commonalities of HGF and H. pylori induced c-Met signalling

ICSB 2008 121 DS2-1-117 Gpd2 remains constantly low. Deleting the GPD1 gene strongly impairs glycerol accumulation while the absence of GPD2 does Temporal perturbation of tyrosine-phosphoproteome not have a pronounced effect. (2) Single deletions of the genes signaling dynamics reveals Src family-mediated global encoding the two 6-phosphofructo-2-kinase isoforms, PFK26 and regulatory networks PFK27 , do not seem to affect the overall glycerol accumulation. Kozuka-Hata, Hiroko1; Oyama, Masaaki1; Tasaki, Shinya2; Semba, Only the double deletion mutant pfk26Δpfk27Δ shows less efficient Kentaro3; Hattori, Seisuke4; Sugano, Sumio2; Inoue, Jun-ichiro1; glycerol accumulation than the wild type. (3) The truncation of Yamamoto, Tadashi1 Fps1 prolongs the phosphorylation period of the MAP kinase 1Institute of Medical Science, University of Tokyo, Tokyo, Japan; Hog1, the core element in the High Osmolarity Glycerol response 2Graduate School of Frontier Sciences, University of Tokyo, Tokyo, pathway. Japan; 3Waseda University, Tokyo, Japan; 4Kitasato University, Conclusions: (1) The roles of the two isoforms of the NAD+- Tokyo, Japan dependent glycerol-3-phosphate dehydrogenase in glycerol accumulation are not equivalent. Mathematical modelling Objective: It is well-known that signal transduction system within indicates a predominant role of Gpd1 in the overall accumulation. a cell leads to the determination of diverse cell fates, such as (2) The two isoforms of the 6-phosphofructo-2-kinase can proliferation, differentiation, or apoptosis. As phosphotyrosine- substitute for each other. The absence of only one of them dependent networks play a key role in transmitting signals, time- does not seem to affect the overall glycerol accumulation. resolved description of their dynamics provides a fundamental Results of mathematical modelling also imply importance of platform for analyzing the regulatory mechanisms at the system 6-phosphofructo-2-kinase in maintenance of energy production level. Here we established a mass spectrometry-based framework in face of increased glycerol demand during osmoadaptation. (3) for analyzing tyrosine-phosphoproteome dynamics through Constitutively open Fps1 impairs the adaptation to hyper-osmotic temporal network perturbation and applied our methodology to stress. the signaling networks that worked in human A431 cells as a model system. DS2-1-119 Results: In this study, we analyzed the dynamics of tyrosine- phosphoproteome upon EGF stimulation using highly sensitive System-level analysis of EGFR signal transduction Dedicated nanoLC-MS/MS system. The time-course activation profiles based on quantitative temporal data of protein tyrosine Posters were generated from the quantitative information on the peptides phosphorylation encoded by the Stable Isotope Labeling by Amino acids in Cell Tasaki, Shinya1; Oyama, Masaaki1; Nagasaki, Masao1; Kozuka- culture (SILAC) technology. Our analysis revealed the property Hata, Hiroko1; Semba, Kentaro2; Gotoh, Noriko1; Hattori, of multi-phase network activation, comprising spike signal Seisuke3; Inoue, Jun-ichiro1; Yamamoto, Tadashi1; Miyano, transmission within 1 min followed by prolonged activation Satoru1; Sugano, Sumio1 of multiple Src-related molecules. Temporal perturbation of 1The University of Tokyo, Tokyo, Japan; 2Waseda University, Src-family kinases with the corresponding inhibitor PP2 in the Tokyo, Japan; 3Kitasato University, Tokyo, Japan prolonged activation phase enabled us to clearly distinguish between sensitive and robust pathways to this treatment, Objective: Epidermal growth factor receptor (EGFR) is a member providing a system-level view of Src function in EGF signaling of of receptor type tyrosine kinases which regulate multiple cellular A431 cells. activities such as growth, differentiation and survival through Conclusions: Our network perturbation strategy based on highly transmitting extracellular signals to intercellular molecules. Binding time-resolved description of tyrosine-phosphoproteome dynamics of ligands to EGFR causes autophosphorylation of several provided a network-wide view of the pathway clusters controlled tyrosine residues in the C-terminal region of EGFR. This tyrosine by Src-family kinases in A431 cells. Our methodology enables us phosphorylation is a key event in initiating downstream EGFR to refine literature-based network structure into cell type-specific signaling. Thus, to elucidate the roles of each autophosphorylation architecture. We expect that mathematical analyses on cell type- site in signal regulation is very important to understand and specific network models will lead us to efficient identification of manipulate EGFR signal transduction system. In this study, we potential drug targets in each disease condition and also enable have analyzed system-level roles of EGFR autophosphorylation us to theoretically estimate the effect of their corresponding drugs site based on quantitative and comprehensive temporal data of on a network-wide scale prior to clinical application. protein tyrosine phosphorylation measured by using LC/MS/MS. Results: From the result of interactome analysis of EGFR DS2-1-118 autophosphorylation sites, variety of proteins that bind to phosphotyrosine 992 (Y992) was most significant among the Mechanisms of glycerol accumulation in saccharomyces analyzed residues. This suggested that Y992 was a functionally cerevisiae upon hyperosmotic stress important autophosphorylation site in regulation of signal Petelenz, Elzbieta1; Kühn, Clemens2; Nordlander, Bodil1; Klein, transduction. Thus, we have focused on Y992 as a model Dagmara3; Jacobson, Therese1; Dahl, Peter1; Schaber, Jörg2; system. To elucidate how Y992 regulates signal network, we Klipp, Edda2; Hohmann, Stefan1 used two NIH3T3 cells that express either wild type EGFR (WT) 1University of Gothenburg, Cell and Molecular Biology, Göteborg, or mutant EGFR with substitution of tyrosine 992 to phenylalanine Sweden; 2Max Planck Institute for Molecular Genetics, (Y992F). We measured temporal activation of protein tyrosine Computational Systems Biology Group, Berlin, Germany; 3Thermo phosphorylation in response to EGF treatment in each cell by Fisher Scientific, Clinical R&D Laboratory - Oxoid, Basingstoke, using Stable Isotope Labeling by Amino acids in Cell culture United Kingdom (SILAC). Next, we constructed functional network among quantified proteins based on Ingenuity Pathways Knowledge Objective: The aim of this work is to establish the relative Base. Then, by mapping the difference of phosphorylation contribution of different mechanisms responsible for amount and temporal pattern to the constructed network, accumulating glycerol as a compatible solute upon hyper- we characterized the influence of mutation on EGF signal osmotic stress in Saccharomyces cerevisiae. We distinguish transduction. three mechanisms of glycerol accumulation: (1) increase in Conclusions: Pathway analysis revealed that temporal pattern glycerol production catalysed by the NAD+-dependent glycerol- alteration of EGFR degradation pathway and Erk1/2 activation 3-phosphate dehydrogenase, (2) stimulation of glycolysis by pathway occurred between WT and Y992F. To estimate what the 6-phosphofructo-2-kinase and (3) preventing the outflow of caused this alteration, we constructed a numerical simulation glycerol present in the cell by closing of the aquaglyceroporin model of EGFR signal transduction and evaluated the possible Fps1. mechanisms by comparison of kinetic parameters. Results: (1) The level of Gpd1 increases rapidly 10 minutes after the osmotic shock; reaches a maximum after 30 minutes and remains constantly high for several hours, whereas the level of

122 ICSB 2008 DS2-1-120 each gene with its respective expression abundance change to evaluate the statistical significance of gene expression in over Spatial information encoded in localized gradients 500 pathways archived in KEGG and BioCarta databases. The generated by signaling cascades aforementioned pathway analysis associated with literature survey Neufeld, Zoltan1; Munoz-Garcia, Javier2; Kholodenko, Boris3 and experimental validation suggest that the uptake of SWCNT 1University College Dublin, Mathematical Sciences, Dublin, into the macrophages is able to activate various transcription Ireland; 2University College Dublin, Dublin, Ireland; 3Thomas factors such as nuclear factor κB (NF-κB) and activator protein Jefferson University, Philadelphia, United States 1 (AP-1) and this leads to oxidative stress, the release of pro- inflammatory cytokines, the recruitment of leukocytes, the Objective: Cascades of covalent protein modification induction of protective and anti-apoptotic gene expression and cycles convey signals from cell-surface receptors to target the activation of T cells. The resulting innate and adaptive immune genes in the nucleus. Each cycle consists of two or more responses may explain the chronic pulmonary inflammation and interconvertible protein forms, for example, a phosphorylated and granuloma formation caused by SWCNT. unphosphorylated protein, and an active, phosphorylated protein Conclusions: The aforementioned pathway scoring method signals down the cascade. has been demonstrated to be a useful tool for microarray-based While signaling cascades were studied experimentally and signaling pathway analysis. A Java-based web server is under theoretically for more than half a century, most studies constructed to provide a user-friendly interface for visualization disregarded the spatial aspects of signal propagation, considering and miming of microarray data at the pathway level. one or more well-mixed compartment(s) with no variation in spatial dimensions. DS2-1-122 Despite important breakthroughs in understanding the input- output relationships and temporal dynamics of information Fibroblast quiescence: Identification of the key processing, we currently lack sufficient theoretical and transcription factors experimental insights into spatial propagation of signals Kondrakhin, Yuriy1; Sharipov, Ruslan2; Filipenko, Maxim3; generated by protein modification cascades. The aim of this Boyarskikh, Ulyana3 work is to develop a quantitative understanding of how activity 1Institute of Systems Biology; Design Technological Institute of gradients spread in space by the subsequent levels of signaling Digital Techniques SB RAS, Novosibirsk, Russian Federation; cascades, how the spreading of phosphorylation signals depends 2Institute of Systems Biology; Institute of Cytology and Genetics on the number of cascade levels and how the gradients of SB RAS, Novosibirsk, Russian Federation; 3Institute of Chemical phosphoproteins along the cascade are controlled by the kinetic Biology and Fundamental Medicine SB RAS, Novosibirsk, Russian properties of the kinases and phosphatases. Federation Posters Results: Using a mathematical model of the signaling cascade Dedicated based on a system of reaction-diffusion equations we analyze the Objective. Regulation of cellular quiescence is critical to spatial activation profiles at different levels of the cascade. We degenerative diseases and cancer. It remains a challenge to show that if the ratio of the kinase and phosphatase activities is identify transcriptional factors (TFs) regulating genes in conditions below a certain threshold, the propagation of the phosphorylation of serum deprivation (SD). We aimed at the two main tasks: signal stalls in space. We demonstrate that a signaling cascade 1) identification of genes overexpressed after SD; 2) promoter produces a set of steady-state activation profiles that for different analysis of identified genes to reveal the TFs regulating them. levels of the cascade decay at different locations. We determine Results. Overexpressed genes were identified by a statistical the conditions and characteristics of the signal propagation method based on optimization of parameters of hyper- and investigate the effect of saturation of the activation and geometrical distribution. Recognition of potential binding sites deactivation reactions. of TFs was done with use of a weight matrix method [2]. Initial Conclusions: The interplay between the reaction cascade frequency matrices were extracted from the TRANSFAC (http:// dynamics and molecular diffusion provides a simple mechanism www.biobase-international.com/pages/index.php?id=transfac) that generates robust positional information, which can be used and JASPAR (http://jaspar.genereg.net). Statistical simulation was by various signaling pathways to regulate a range of cellular applied to decrease the number of false positive results. processes. Analyzing the microarray data, we selected two groups of genes: 1) containing the genes significantly overexpressed after SD; DS2-1-121 2) containing the genes with not changed expression (control group). Sequences of promoter [-1000bp, +1000bp]-regions of Microarray-based pathway analysis of single-walled carbon all selected genes were extracted from Ensembl (http://www. nanotubes induced pulmonary cytotoxicity ensembl.org). Promoter regions of genes from both groups Chou, Cheng-Chung were analyzed by the matrix method for recognition of TFs National Chung Cheng University, Department of Life Science, binding sites. Comparison of the occurrence frequencies of Chia-Yi, Taiwan recognized binding sites allowed to reveal TFs responded to serum deprivation (for instance, KLF4, PPARG and FOXF2). Objective: Carbon nanotubes are a nanomaterial that is Application of chi-squared independence test allowed us to reveal extensively used in industry. The potential health risk of additional set of TFs, whose binding sites significantly co-occured chronic carbon nanotubes exposure has been raised as of in promoters of the genes overexpressed in SD conditions. The great public concern. It is well known that one of the common most striking examples were IRF2, c-Ets-1, PU1 and STAT6. cytotoxic responses to SWCNT exposure in lung is granuloma Conclusions. Three TFs KLF4, PPARG and FOXF2 may formation. The granulomas are mainly composed of aggregates orchestrate quiescence program in fibroblasts. Their potential of macrophages laden with SWCNT particles and thereof an coregulators were also revealed. investigation of the genome-wide gene expression changes in Acknowledgements. SB RAS Integrational Grant n13 “Stem macrophages exposed to SWCNT, associated with a detailed cells for future biotechnology”. signaling pathway analysis, ought to provide molecular insights References. into this type of granuloma formation. 1.H.Liu et al. (2007) A transcriptional program mediating entry into Results: In the present study, we used Affymetrix microarrays cellular quiescence, PloS Genetics, 3:e91:1-13. to investigate the genome-wide effects of SWCNT on human 2.E.A.Ananko et al. (2007) Recognition of interferon-inducible macrophage-like cells differentiated from a human monocytic sites, promoters, and enhancers, BMC Bioinformatics, 8:56:1-14. leukemia cell line THP-1 at the molecular level. To extract meaningful biological knowledge from huge microarray data, we developed a computational method to evaluate and select the pathways that are most affected by transcriptional changes in genome-wide gene expression experiments. The method weights

ICSB 2008 123 DS2-1-123 each TF and identifies the TFBSs that are with high probability responsible for the expression regulation. By intersecting high- NCoR activity in the prostate Ssystem confidence target TFBSs with known SNPs we identify sets Sebastiano, Battaglia1; James, Thorne2; Chris, Bunce3; Moray, of SNPs that are likely to affect gene expression in human Campbell4 populations. 1University of Birmingham, Medical School, Birmingham, United Kingdom; 2University of Birmingham, Medical School, DS2-1-125 Birmingham, United Kingdom; 3University of Birmingham, School of Biosciences, Birmingham, United Kingdom; 4Roswell Park Integrated model of apoptosis Cancer Institute, Dept. of Pharmacology and Therapeutics, Likhovidova, Elena; Sharipov, Ruslan; Kolpakov, Fedor Buffalo - NY, United States Institute of Systems Biology, Design Technological Institute of Digital Techniques, Novosibirsk, Russian Federation Prostate cancer is one of the most common malignancies and second cause of death in men over-50. Non-malignant prostate Objective: Formal description of pro- and anti-apoptotic epithelial cells (RWPE-1) display sensitivity to a wide range of small machinery was enriched significantly for the last years, and a lipophilic molecules that act as ligands for nuclear receptors (NRs) range of models was created. But, mainly, they describe different to regulate target genes. By contrast, prostate malignant cells (PC- segments of implicated pathways, and the total jigsaw puzzle - 3) display suppressed responses. NCoR-1 (Nuclear CoRepressor-1) model of apoptosis regulation - has not been done yet. The main is a key negative regulator of NRs and is up-regulated in prostate goal of this work was combining of individual models to create the malignancies. We designed three different shRNA directed against comprehensive model of pro- and anti-apoptotic pathways for in exon 5, 38 and 44 of NCoR-1 in order to create a stable knock- silico experiments. down in PC-3 cells. Cell viability assays demonstrated increased Results: Nine individual models published for the last four years sensitivity to several NR ligands, notably PPARs. and describing different parts of pro-/anti-apoptotic pathways A microfluidic qRT-PCR approach was undertaken measuring 96 were reconstructed using BioUML workbench [1] and deposited genes targets simultaneously and quantitatively. These included at the BMOND database [2] in category «Apoptosis». Three NRs and NRs co-factors, histone-modifying enzymes (e.g. models [3-5] were used for construction of the integrated model, Dedicated HDACs, CPB, PADI4), PPAR and VDR target genes (GADD45, which comprises 61 proteins and their complexes and 34 Posters CYP3A4, CYP24, IGFBP3), cell cycle controls (CDKN1A, TP53) reactions. This model describes induction of apoptosis by TNF-α and metabolite handling components (e.g. PTGS2, TGF-beta, via TNFR-1 receptor. TNF-α also activates transcription factor CYP27b1, CYP19). RWPE-1, PC-3, shNCoR-PC-3 and VO-PC-3 NF-κB inducing cell survival (e.g., via caspase inhibitor XIAP). cells were FACS sorted in G1, S and G2, RNA extracted and gene As the next step of model development, selection and analysis expression analysis performed. SAM (Significance Analysis of of new experimental data on apoptosis was started for further Microarray) showed 66 genes significantly differentially expressed enhancement of the integrated model. between PC-3 and RWPE-1 cells in a cell cycle dependent Conclusions: Integration of individual models on apoptosis manner including CDKN1A, CCND1, G0S2, GADD45A, HDACs, regulation was started and the integrated model was assembled. PPARs and VDR. NCoR-1 knockdown significantly up-regulated Enhancement of the model on the base of experimental data is expression of 8 genes: CASP-4, AKR1C1, MYB, ABCG-2, ALOX- performed. 5, TP53, CDKN1A, PPAR-gamma. Subsequent genome wide References: analyses of targeting PPARs with chemical inhibition of NCoR-1 1. BioUML workbench - http://www.biouml.org (with an HDACs inhibitor) demonstrated restored expression of 2. BMOND - Biological MOdels aNd Diagrams Database - PPARs regulated genes (e.g. CDKN1A, TGFBRAP-1). These http://www.bmond.org studies are the first to use a micro-fluidic qRT-PCR and micro- 3. Bagci et al. (2006) Bistability in Apoptosis: Roles of Bax, array approaches to analyze the prostate system confirming the Bcl-2 and Mitochondrial Permeability Transition Pores. key role of NCoR-1 as a central hub in the NR network. Our aim Biophysical Journal. 90:546-1559. is to build a model of NCoR-1 network understanding its precise 4. Rangamani and Sirovich (2007) Survival and Apoptotic role in prostate malignancies. Pathways Initiated By TNF-alpha: Modeling and Predictions. Biotechnology and Bioengineering. 97(5):1216-1229. DS2-1-124 5. Legewie et al. (2006) Mathematical Modeling Identifies Inhibitors of Apoptosis as Mediators of Positive Feedback A systems approach for quantifying the dynamics of and Bistability. PLoS Computational Biology, 2(9), transcription regulatory interactions in higher eukaryotes e120:1061-1073. Balwierz, Piotr; Pachkov, Mikhail; van Nimwegen, Erik; Zavolan, Mihaela DS2-1-126 Biozentrum, University of Basel, Basel, Switzerland Structural robustness analysis of the Huang-Ferrell model Using deep sequencing data of 5’ ends of mRNAs (CAGE of MAPK signaling tags) we identified promoters genome-wide at single base-pair Trané, Camilla1; Jacobsen, Elling2 resolution in the human genome, and quantified their expression 1KTH, Automatic Control, Stockholm, Sweden; 2KTH, Stockholm, across 56 different tissues and conditions. We developed a novel Sweden computational method that combines comparative genomic analysis of proximal promoter sequences with modeling of Objective: Models of biochemical networks are characterized by their expression dynamics to infer time- and tissue-dependent high levels of parametric and structural uncertainty, manifesting transcription regulatory interactions genome-wide. both changes occurring in the real system and incomplete Using state-of-the-art comparative genomic computational knowledge of the biochemical processes. Correspondingly, both predictions of transcription factor binding sites (TFBSs) we show parametric perturbations and structural perturbations should that binding sites for a large number of transcription factors be considered when analyzing the fragility and robustness of (TFs) have very specific positional preferences with respect to a model. Here a method to analyze structural robustness is transcription start sites (TSSs). proposed and applied to a model of the mitogen activated Our analysis infers activities of over 200 regulatory motifs across protein kinase (MAPK) signaling cascade [Huang and Ferrell 56 CAGE libraries, 79 samples of the GNF atlas, and 59 cancer (1996)]. Through extensive parameter searches, Quai et al. cell lines (NCI-60 data set), thereby globally quantifying dynamic (2007) showed that the Huang-Ferrell model structure not only changes in transcription regulatory interactions across different results in an ultrasensitive input-output response, but is capable tissues, disease states, and developmental processes in human. of displaying bistability and sustained oscillations for specific In addition, the modeling of promoter expression profiles in terms parameter combinations. We analyze the nominal Huang- of motif activities identifies high confidence target promoters for Ferrell model to determine how large structural perturbations

124 ICSB 2008 are required to translate the ultrasensitive response curve into DS2-2-11 a bistable and an oscillatory response curve, respectively. This provides a quantification of the robustness of the qualitative Modeling maintenance of the mid-hindbrain boundary stationary behavior of the Huang-Ferrell model towards structural based on qualitative patterning data uncertainty. Wittmann, Dominik1; Bloechl, Florian1; Prakash, Nilima2; Results: By explicitly perturbing the network structure, it is Truembach, Dietrich2; Wurst, Wolfgang2; Theis, Fabian J.3 revealed that the Huang-Ferrell model is highly fragile towards 1HelmholtzZentrum München, German Research Center for structural perturbations. Small structural perturbations can Environmental Health, Institute for Bioinformatics and Systems translate the ultrasensitive sigmoidal response curve into a Biology, Munich, Germany; 2HelmholtzZentrum München, bistable switch, by inducing a saddle-node bifurcation, or result in German Research Center for Environmental Health, Institute of sustained oscillations, by inducing a Hopf-bifurcation. Developmental Genetics, Munich, Germany; 3HelmholtzZentrum Conclusions: Parametric perturbations are routinely used for München, German Research Center for Environmental Health / the purpose of model validation. Robustness towards parametric BCCN - MPI D&S (Göttingen), Munich, Germany perturbations, however, does not imply robustness towards structural perturbations. Nevertheless, structural perturbations are Objective: Shortly after gastrulation the vertebrate neural tube rarely addressed as a model validation tool. Considering structural is patterned along the anterior-posterior axis into four regions, perturbations, the displayed poor robustness of the Huang-Ferrell which continue to develop into forebrain, midbrain, hindbrain and model of MAPK signaling may be a weakness of the model, or a spinal cord. This patterning is induced and determined by a well feature of the real biological system. defined and locally restricted expression of genes. A key player in this process is the isthmic organizer located at the mid-hindbrain Dedicated session 2-2: boundary (MHB), which controls development of mid- and hindbrain. A great deal of experimental effort has been made to Modelling approaches understand the induction and maintenance of the MHB. Here we aim at developing a multi-cellular model of a regulatory network DS2-2-09 that explains the maintenance of this system. Results: First we develop a boolean model based on information Robustness of simple biochemical oscillators depends on from in-situ hybridization experiments. In an exhaustive feedback ‘wiring’ computational analysis we check a total of 531,441 possible Ibig, Ariane; Stelling, Joerg topologies if they maintain the gene expression pattern observed ETH Zurich, Computer Science, Zurich, Switzerland around the boundary. We find that among them only one

regulatory network and some subnetworks obtained by deletion Posters Objective: In theory, the output signal of a cellular control circuit of one or two links have this gene expression pattern as a steady Dedicated can be produced in many different ways; by different wiring of state. The strongly connected network agrees well with models its components. In reality, only a few different topologies ever inferred from loss- and gain-of-function experiments and can arise. It is generally accepted that this is due to their structure partially be validated by a promotor analysis. We then build up offering more robust behaviour for critical parameters. Yet, the dynamic ODE models which reproduce the boolean fixed point dependencies between robustness and wiring are still unclear. over a wide range of parameters. To further explore differences We introduce geometrical tools to analyse surface points of the between the regulatory networks we simulate perturbations of parameter space of biochemical oscillators. The thereby revealed the steady state. We observe that the more densely connected a dependencies between feedback wiring and robustness of network is, the less misplaced gene expression occurs. parameters are exemplified on a simple biochemical oscillator. Conclusions: We demonstrate that the basic structure of the Results: Different geometrical tools determine either properties regulatory network underlying the maintenance of the MHB can of the entire oscillation space or of the robustness of a be recovered from in-situ hybridization images alone. Our analysis single parameter. The volume of the oscillation space can be of this network shows that it is optimized for robustness against approximated by the volume of the convex hull of its surface random deletion of links and misplaced gene expression. points. While the closeness to the center of mass indicates robustness for the concerned parameters, the minimal angle of DS2-2-12 vertices reveals sensitive areas of the oscillation space. By doing this for all possible (feedback and feedforward) wirings of a simple Elongation dynamics shape bursty transcription and oscillator, the high impact of the wiring of the feedback loops translation on the geometrical properties of the oscillation space is evident. Dobrzynski, Maciej; Bruggeman, Frank As a largely global method, however, it does not yield a direct Centre for Mathematics and Computer Science (CWI), dependency between the robustness of a single parameter and Amsterdam, Netherlands the wiring of the oscillator. Therefore, different statistical tools were used to analyze the oscillation intervals of single parameters Objective: Time-averaged stochastic measures such as mean over all wirings. This revealed that every parameter has wirings and noise loose applicability when they require a time-averaging with positive as well as wirings with negative effect on its span much longer than a lifetime of a single cell. Then, the robustness and that different parameters do not necessarily have waiting time statistics for particular processes become an similar robustness properties for the same wiring. important determinant of the stochasticity in cell physiology. Conclusions: High-dimensional geometrical methods can be Bursts in synthetic activity have a great potential for generating used to analyze robustness properties of biological oscillators, this stochasticity and resulting cellular heterogeneity. Bursts based on the surface points of their parameter space. Combined are characterized by periods of high activity followed by silence with our recently developed algorithm to compute the needed periods causing different cells to experience bursts varying in size, surface points, analysis of more complex biochemical oscillators duration and timing. We present an analysis of the stochastics of (ultimately the circadian clock) can lead to insights into parameter interarrival times of production events, originally developed for the dependencies. analysis of telecommunication networks. We show its relevance for the analysis of transcriptional and translational elongation dynamics. Results: We define general measures for characterization of bursts (burst size, significance, and duration) from experimental data. The discovery of bursts in mRNA and protein production by others inspired us to use those indices to determine burst properties of stochastic motion of motor proteins along biopolymer chains. Using theoretical models, we find that bursts at the input of the chain tend to be smoothed out by longer

ICSB 2008 125 chains due to congestion of motor proteins. At a fixed initiation Forward Loop (C1-FFL) and Incoherent type-1 Feed-Forward rate, bursts can emerge due to pausing of motor proteins. If Loop (I1-FFL) are used to concern how the yeast respond to protein production takes place from independent bursty sources, different environmental changes. To identify the genetic networks we show that the resultant bursts loose their significance as the in response to seven environmental stresses (i.e. heat shock, number of sources increases. hydrogen peroxide, Menadione, Diamide, Hypo-osmotic shock, Conclusions: Our analysis indicates two possibilities for DTT and Sorbitol osmotic shock), microarray data, ChIP-chip generating bursts in gene expression. Strong repressors data and protein-protein interaction (PPI) data are incorporated could induce bursts according to the bursty initiation. Highly- into our genetic dynamic models which describe the dynamical

activated genes can give rise to bursts by means of the interplaying behaviors. The models with the transcription level ai(t)

pausing mechanism. Pausing has been well documented for and the protein level ei(t) of the i-th gene are formulated as: äi(t)+ci

RNA polymerase and ribosomes. Recent experiments confirm ài(t)+diai(t)=∑j=1fijej(t-τj)+fi0+εi (1). èi(t)=αiai(t)-λiei(t)-∑j=1gijei(t)ej(t)+gi0+ξi, the increase of noise in gene expression as a result of RNA i=1,..N (2). Eq.(1) and eq.(2) denote transcriptional regulations and polymerase pausing. translational regulations, respectively. Akaike Information Criterion method is exploited to identify the significant transcriptional

DS2-2-13 regulation parameters fij and protein interaction parameters gij to construct genetic regulatory dynamics. The genetic regulatory

Using gene network dynamics to obtain a holistic view PPI networks could be constructed through linking fij and gij, on cell-cell communication in the pancreas tumor respectively. microenvironment Results: From the identified networks, we find a total of 12 Busch, Hauke1; Rogon, Zbigniew1; Szabowski, Axel2; Eils, SIMs which are proposed to stabilize the networks and a Roland1; Giese, Nathalia3 total of 140 C1-FFLs and 164 I1-FFLs which are proposed to 1German Cancer Research Center, Theoretical Bioinformatics, activate biological functions over a short and long period of time, Heidelberg, Germany; 2German Cancer Research Center, respectively. Heidelberg, Germany; 3University of Heidelberg, Hospital, Conclusions: With experimental evidence, the ability of CIN5 Department of General Surgery, Heidelberg, Germany affects network stability under the stresses with the order Hypo-osmotic shock>heat shock>Diamide>hydrogen peroxide. Dedicated Objective: Translation of large-scale omic data into a coherent Besides, we also find that four biological functions, ENERGY, Posters model for cellular regulation allowing to simulate, predict and PROTEIN FATE, REGULATION OF METABOLISM AND PROTEIN control cellular behavior is far from being resolved. Global FUNCTION, and TRANSPOSABLE ELEMENTS, VIRAL AND regulation of cell homeostasis and cell fate requires a complex PLASMID PROTEINS, are activated over a short period of time and controlled interplay of protein signaling and gene regulation, (pulse) by stress crosstalk through I1-FFL, and one biological Combining these processes into one model is inherently difficult, function, CELL TYPE DIFFERENTIATION, is activated over a long as they occur on different time scales in the range of minutes to period of time by stress crosstalk through C1-FFL. hours, respectively. We propose a complexity reduction approach to capture cell-fate decisions based on the slaving principle, DS2-2-15 which states that the long-term macroscopic behavior of a system is controlled by its slowest evolving variables. In biological Simulation and metabolome analysis of human red blood terms this means that long-term phenotypic behavior of a cell is cells during storage in MAP medium at 4 °C reflected in its gene expression kinetics. Nishino, Taiko1; Yachie-Kinoshita, Ayako2; Suematsu, Makoto2; Results: Tumor microenvironments are established by paracrine Tomita, Masaru1 cell-cell communication via cytokines taken up and secreted 1Institute for Advanced Biosciences, Keio University, Fujisawa, by tumor and stroma cells. While playing a crucial role in Japan; 2School of Medicine, Keio University, Biochemistry and defining tumor progression and malignancy, microenvironments Integrated Medical Biology, Tokyo, Japan likewise constitute easily accessible drug targets, namely the genetically stable stroma cells and defining cytokine patterns. Objective: Improving red blood cell (RBC or RC) storage We unraveled the paracrine communication of pancreatic tumor system using Mannitol-adenine-phosphate (MAP) solution, with surrounding stellate cells by individually stimulating cells which is widely used, has been an important issue especially in with conditioned supernatant from the respective other cell type. emergency medicine. During storage, RBCs progressively lose Recording gene expression profiles time-resolved up to 24 hours their viabilities and functions by reducing intracellular ATP and after stimulation identified key genes based on their kinetic profile 2,3-bisphosphoglycerate (2,3-BPG). These two metabolites as well as their biological function. Using a genetic algorithm are used as indicators for assessment of the expiration date of combined with a search for robust system solutions, we reverse RC-MAP. However, the detailed mechanism of this metabolic engineered an dynamic gene regulatory network describing failure in stored RBC remains unclear. In this study, to propose the cytokine-induced response in cell homeostasis. The model possible candidates of the target for novel storage method which predicts points of interference modulating of interruption the can retain the amount of ATP and 2,3-BPG, we reconstruct tumor microenvironment communication, which are verified in the MAP-stored condition at low temperature and predict the vitro and now await validation in situ. metabolic behavior of RC-MAP using mathematical model of Conclusions: Our gene expression kinetics analysis and human erythrocyte covering wide range of metabolic pathways, modeling approach provide a novel way of obtaining a cell-wide connection between hemoglobin and metabolism. (Kinoshita, A. view on the dynamic orchestration of cellular pathways and a et al., The Journal of Biological Chemistry, 282, 10731-10741. broad degree of interdependency that control cell-fate decisions. (2007)) Results: To characterize the MAP-stored conditions, we set DS2-2-14 five external parameters including pH variation during storage, inhibition of enzyme activities and Na+/K+ pump activities by low Identifying network motifs and crosstalk of environmental temperature, activation of enzymes in purine metabolism by stress-activated genetic networks via microarray data lowed pH, and stabilization of hemoglobin state to R-state by low Li, Cheng-Wei1; Chen, Bor-Sen2 temperature. As a result, the time-courses of ATP and 2,3-BPG 1Institute of of Electrical Engineering, Hsinchu, Taiwan; 2National in MAP-stored RBC predicted by the above model showed good Tsing Hua University, Department of Electrical Engineering, agreement with those of previously reported experimental data. Hsinchu, Taiwan Further analysis of the model suggested that the five external parameters should be strongly considered to preserve the quality Objective:Patterns that occur in a real genetic network of stored blood. We are now validating and refinement of the RC- significantly more often are called network motifs, which must MAP model by comparing with metabolome data collected using have been preserved over evolutionary timescales against capillary electrophoresis / time-of-flight mass spectrometry (CE- mutations. Single Input Module (SIM), Coherent type-1 Feed- TOFMS).

126 ICSB 2008 Conclusions:This work indicates the possibility of the between the original and the resulting perturbed system is mathematical model of large-scale metabolism in human RBCs to calculated. The state giving the smallest error norm is chosen represent the metabolic status in cold-MAP-stored RBCs and to to be collapsed at the end of the current iteration. This process predict the way to maintain its quality efficiently. is repeated until the error norm between the original and the reduced system is not tolerable. This iterative approach produces DS2-2-16 an ordered list of states to be collapsed based on the worst-case error they would result in, and also provides an indirect route to Soft integration of data for large-scale reverse engineering reduce the nonlinear system with an estimated error between the Gustafsson, Mika1; Hornquist, Michael1; Bjorkegren, Johan2; original and reduced systems. Tegnér, Jesper3 Conclusions: We have applied this method to a biological 1Linkoping University, Department of Science and Technology, network based on the bacterial EnvZ-OmpR two-component Norrkoping, Sweden; 2Karolinska Universitetssjukhuset, system. The nonlinear system model is reduced from 20 to 4 Department of Medicine, Stockholm, Sweden; 3Linkoping states which capture the essential structure and features of the University, Division of Computational Biology, Linkoping, Sweden original network. Both simulation studies and sensitivity analyses show that the reduced and original models behave approximately Objective: We present a methodology to reverse engineer within the tolerance, and hence support the theory. Gene Regulatory Networks (GRNs) by integration of data Reference: from microarrays, TF-binding experiments, protein-protein [1] Miles S. Okino and Michael L. Mavrovouniotis, Simplification of interactions, CAGE, text-mining, etc., within an ODE framework. Mathematical Models of Chemical Reaction Systems, Chemical The methodology is generally applicable and based on Least Reviews, 1998, Vol. 98, No. 2, p391-408. Absolute Deviations (LAD), which is the maximum likelihood estimator for the broad Laplace error distribution normally present DS2-2-19 for microarrays. The approach is fully data driven and handles both non-linearities and data integration in a “soft evidence” way. A comprehensive model of cellular decision making in S. The integration fuses expression data sets with structural a priori cerevisiae knowledge, where each experiment type is relied upon only to the Waltermann, Christian1; Klipp, Edda2 extent it lowers the prediction error. 1Max Planck Institute for Molecular Genetics, Berlin, Germany; The reverse engineering procedure utilizes a weighted version of 2Max Planck Institute for Molecular Genetics, Computational LAD, combined with a weighted L1-shrinkage of the predictor Systems Biology Group, Berlin, Germany coefficients. The algorithm follows a solution by gradually increasing the L1-norm of the predictor coefficients, producing Objective: In their natural environment, cells are exposed to a Posters optimal solutions for all possible values of the shrinkage wide range of stresses ranging from changes in osmolarity and Dedicated parameters in a single run. The implementation is based on the temperature to alterations in nutrient availability. In S. cerevisiae KKT-conditions, which enables ultra fast execution time in the (budding yeast) sensed environmental variations are mediated case where the number of predictors vastly exceeds the number by a complex signaling network and lead to adaptation through of experiments. modification of the transcriptome as well as through cell cycle Results: The proposed methodology is evaluated by a case regulation. The traditional view of signaling networks often study of human cells (THP-1) undergoing differentiation by restricts the analysis to one or two pathways at a time, thereby stimulation with PMA. The data is recently obtained from the neglecting the interaction between multiple sources of stress RIKEN consortium and consists of three types of expression data acting on the cell. The availability of experimental data and (qRT-PCR, CAGE and Illumina arrays), comprising a total of 23 parameters in numerous signal transduction pathways in yeast experiments and almost 1700 transcription factors. As a result, makes it however feasible to take a broader view on cellular we are able to fuse the qRT-PCR data together with the CAGE signaling and decision making. and Illumina data sets and several types of binding data, to obtain Results: Here we present a mathematical model of a large-scale transcriptional gene regulatory network within a interconnected pathways in S. cerevisiae and their effect on the limited amount of cpu-power. cell cycle, which is used to study a combination of stressors Conclusions: Here we present one possible practical solution to acting on budding yeast cells. Read-outs of the model include the the large-scale data integration problem to include as much of the most likely responses of a cell when subjected to a combination biological data material as possible. of stimuli. Conclusions: Our model predicts the common types of cell DS2-2-18 cycle regulation under nutrient limiting or stress-rich conditions. Results from in vivo robustness assays (gTOW) will be used to A systematic, algorithmic methodology for reducing discriminate between different model variants. biochemical network models Chang, Yo-Cheng1; Papachristodoulou, Antonis2 DS2-2-20 1Oxford Centre for Integrative Systems Biology, Department of Biochemistry, University of Oxford, Oxford, United Kingdom; FAD-synthetases and riboflavin production by yeasts 2Control Group, Department of Engineering Science, University of Sydorovych, Inna Oxford, Oxford, United Kingdom Institute of Cell Biology, Lviv, Ukraine

Objective: Model reduction is a commonly used technique for Objective: Some yeast species capable to riboflavin analyzing and understanding large-scale biological networks oversynthesis in iron-deficient medium were described as which work in multiple time- and space- scales. Singular flavinogenic. Activities of enzymes involved in flavins biosynthesis perturbation, a traditional model reduction technique, relies on in these yeast species are regulated by iron except of riboflavin prior knowledge to identify true fast and slow reactions so that the kinase and FAD-synthetase which catalize the last steps in flavins original model can be reduced correctly [1]. However, in most real biogenesis. cases in biology, such prior knowledge may not exist. In this work Results: Phylogenetic analysis of enzymes that participated in we have developed a systematic, algorithmic methodology to flavins biosynthesis was performed. It was shown that primary quantify the worst-case difference in the behaviour between the structure of FAD-synthetases of yeasts which for their riboflavin original and reduced systems based on control theory tools. At production in iron deficient medium distinguish is quite different. the same time, this method provides a priority list of states to be 2 protein-protein interactions were reported for FAD-synthetases collapsed depending on their effect on this error. (http://dip.doe-mbi.ucla.edu/dip/). One of them - the interaction Results: The original nonlinear system is first linearized about with h-subunit of ATP-synthase and the other – with Soh1p. a steady-state. Then, in every iteration, each candidate state Soh1p is a conservative component of Mediator complex which is collapsed by singular perturbation, and the norm of the error interacts with activators or repressors of transcription and with

ICSB 2008 127 RNA polymerase II. Microarray experiments in SPELL database of various classes of ion channels, and electrochemical processes, (http://imperio.princeton.edu:3000/yeast ) were used for design into a single overarching model. This model elucidates how even a of interactions between Soh1p and Fad1p. It was studied the relatively small number of ion channels (~ 5 MScL) can guarantee conformity of transcription repression of FAD1 and SOH1 genes a reliable osmoregulation; and how a relatively constant proportion with expression of RIB1 gene that codes for the first enzyme of a population of E.coli survives osmotic shocks, even without of flavinogenesis - GTP-cyclohydrolase II. Repression of both mechanosensitive ion channels. The model is part of a larger FAD1 and SOH1 genes causes the reduction of RIB1 expression model of bacterial responses to multiple stresses. (microarrays by Causton 2001, Gross 2000). At the same time Results: Our model describes the response of bacterial cells increased expression of RIB1 is observed only in case of high to osmotic shocks. It shows that it is due to a delicately tuned level expression of both FAD1 and SOH1 genes (microarray mechanism that E. coli can survive even dramatic environmental experiments by Duvel 2003, Leber 2004). changes. It also suggests how stochastic gene expression might Conclusions: Based on such analysis we suppose that be beneficial for the growth of a bacterial colony and how bacteria FAD-synthetase might be of special importance for riboflavin may cope with multiple stresses. oversynthesis in iron deficient medium. Conclusions: A multitude of interlaced processes guarantees a robust osmoregulation in bacterial cells. DS2-2-22 DS2-2-24 A system-level study on the dynamical characteristics of E. coli swarming Role of Na+-Ca2+ exchange in neonatal and adult ventricular Kim, Tae-Hwan1; Baek, Songjoon2; Jung, Sung Hoon3; Cho, cells: a simulation study Kwang-Hyun1 Sano, Hitomi; Naito, Yasuhiro; Tomita, Masaru 1Korea Advanced Institute of Science and Technology (KAIST), Keio University, Institute for Advanced Biosciences, Kanagawa, Department of Bio and Brain Engineering, Daejeon, Republic Japan of Korea; 2National Center for Toxicological Research, USFDA, Division of Personalized Nutrition and Medicine, Arizona, Objective: The present paper expanded an integrated United States; 3Hansung University, Dept. of Information and mathematical model, which successfully simulated developmental Dedicated Communications Engr., Seoul, Republic of Korea changes in electrophysiological activities of rodent ventricular Posters cells, to contrive models that reflect a hypothesis on the role Objective: Chemotaxis is a movement control mechanism of Na+-Ca2+ exchange in neonatal ventricular cell: the Na+-Ca2+ of bacteria which enables them to detect certain chemicals in exchange assumes a relatively greater role in neonatal cells with a their environment and to induce an appropriate response. For morphologically sparse sarcoplasmis reticulum (SR). Results: Our E. coli, this is achieved through controlling tumbling frequency simulation with the expanded model on the basis of the E-CELL that is determined by the rotational direction of flagella-attached Simulation Environment (SE) showed that neonatal ventricular rotary motors depending on the local gradient of attractants or cell consumed larger amount of adenosine triphosphate (ATP) repellents. Overall, their motile behavior exhibits biased random than adult ventricular cell did, owing to long relaxation time. The walks with a tendency to swim towards (away) specific attractants amount of ATP consumed per beat was reduced to the amount (repellents). Interestingly, they also exhibit a kind of social behavior equivalent to adult ventricular cell, as increasing the relative current + 2+ and many experiments and simulation studies have been carried density of Na -Ca exchange (FNCX) by 4-fold, which is consistent out to study such phenomena. However, the fundamental with the observed current density in neonatal rabbit ventricular cell. characteristics of E. coli swarming behavior are still largely To further validate the hypothesis, the relative functional activity

unknown. The aim of our study is to unravel the fundamental of SR (FSR) was incrementally increased from neonatal to adult

characteristics at a system-level. levels and subsequently increased FNCX, and found that increasing

Results: From extensive computer simulations under various FNCX had negative impact on sarcomere relaxation velocity in

environmental conditions, we found that the dispersion of E. coli ventricular cell model with large FSR. Conclusions: The role of follows the Taylor’s power law and shows a negative binomial large Na+-Ca2+ exchange current density in neonatal ventricular distribution regardless of environmental topology. Moreover, cell is to accelerate the relaxation of sarcomere by increasing we found that the distribution of attractants follows the Pareto transsarcolemmal Ca2+ efflux and shortening the Ca2+ transient distribution which is a power law distribution found in many types lasting time. The over-representation of Na+-Ca2+ exchange in of clustering behavior including social, scientific, and geophysical neonatal ventricular cell thus plays an important role as substitute phenomena. for immature SR Ca2+ ATPase (SERCA) that reuptakes Ca2+, Conclusion: From our analysis, it was revealed that the because over-representation of SERCA consequently results in distribution characteristics of E. coli and attractants are primarily larger amount of ATP consumed per beat. determined by the tumbling frequency of E. coli. This suggests that the distribution of E. coli can be controlled by changing the DS2-2-25 tumbling frequency. Moreover, it also implies that the tumbling frequency might be used as a discrimination factor among Expanding the dimensions in cellular modelling: An different microorganisms. integrated model of metabolism, transcription, and protein synthesis in E. coli DS2-2-23 Thiele, Ines; Bordbar, Aarash; Fleming, Ronan M.T.; Palsson, Bernhard O. Modeling homeostasis of turgor pressure in E. coli University of California, San Diego, Bioengineering, La Jolla, Thiel, Marco1; Romano, M. Carmen1; Grebogi, Celso1; Booth, United States Ian2; Miller, Samantha2 1University of Aberdeen, Department of Physics, Aberdeen, Objective: The reconstructions of biological networks is United Kingdom; 2University of Aberdeen, Institute of Medical a common denominator in systems biology. In particular, Sciences, Aberdeen, United Kingdom genome-scale metabolic reconstructions have proven useful for biological discovery, metabolic engineering, and phenotypic Objective: Effective osmoregulation is crucial for the survival characterization. However, these reconstructed networks omitting of bacteria. Mechanosensitive ion channels have shown to other cellular functions such as synthesis of macromolecules be essential for the cell to survive osmotic shock. We present such as RNA and proteins, and hence neglecting important a model for the gating of mechanosensitive channels, which determinants of cellular energy requirements and biomass yield. successfully describes both osmotic up- and downshock. The Results: We reconstructed a comprehensive reaction network most challenging problem for the modeling is the combination for RNA and protein synthesis accounting more than 1500 E. of many different aspects of the regulation process, e.g., the coli genes (~ 35% of E. coli’s genome). All network reactions biophysical characteristics of the cell wall, the combined activation were formulated as sequence-specific and mass-balanced

128 ICSB 2008 interconversions. We considered all known transformations DS2-2-28 necessary to produce the functional gene products, rendering this reconstruction to be one of the most comprehensive A strategy for assessing sensitivities in biochemical models databases on transcription and translation in E. coli. The resulting with uncertain parameter values reconstruction was used to determine the cost of the protein Sahle, Sven1; Mendes, Pedro2; Hoops, Stefan3; Kummer, Ursula1 synthesis machinery converting it into a mathematical format 1University of Heidelberg, Modeling of biological processes, and carrying out constraint-based analysis. The simulation Heidelberg, Germany; 2University of Manchester, Manchester results convincingly demonstrated the significant energy costs Centre for Integrative Systems Biology, Manchester, United associated with the transcription process. Previously published Kingdom; 3Virginia Tech, Virginia Bioinformatics Institute, work has suggested that many genes are only expressed once Blacksburg, United States per cell cycle. The computed results strongly support that this observation is directly correlated with energetic costs of Objective: When modeling biochemical systems one of the major transcription. Furthermore, we integrated the synthesis network problems is usually incomplete knowledge about the values of with a metabolic network of E. coli to demonstrate the effects the kinetic parameters. Often some kind of sensitivity analysis, of different environmental conditions on growth performance, such as metabolic control analysis, is employed to quantify the ribosome production capacity, and ATP yield. The direct importance of each of the model’s parameters for the behaviour consideration of the RNA and protein synthesis machinery as part of the system. These commonly used sensitivity analysis methods of biomass enabled quantitative predictions from genotype to provide only local information, i.e. they give the importance of a phenotype under various physiological genetic conditions. parameter provided all the parameter values are already known. Conclusion: This work represents a significant step towards An alternative approach is global sensitivity analysis, where the more comprehensive and mechanistically detailed, cellular sensitivities are evaluated for a whole range of parameter values, models to study the synergy of metabolism and macromolecular usually either by scanning the parameters or by using random biosynthesis. These models will be particularly relevant for sampling, both of which are extremely expensive computationally metabolic engineering and for investigating biological objectives of if the parameter space is high dimensional. Here, we will living cells. present a more efficient approach employing a range of different optimization methods. DS2-2-26 Results: Using the software tool COPASI we apply several optimization algorithms to the flux control coefficients of models Model identification in systems biology: Solving ill-posed available from the biomodels.net database. With this we can problems using regularization efficiently estimate minimal and maximal control of different 1 1 2 Mueller, Stefan ; Lu, James ; Machne, Rainer reaction steps for a whole range of parameters. We discuss which Posters 1Radon Institute for Computational and Applied Mathematics, optimization algorithms are especially suitable for this approach. Dedicated Austrian Academy of Sciences, Linz, Austria; 2Theoretical Conclusions: We present a technique that can be used to Biochemistry Group, University of Vienna, Vienna, Austria analyze the sensitivity coefficients of a model over a large and high dimensional parameter range in a computationally efficient Objective: The quality of a mathematical model for a biological way. This allows the analysis of the global flux control properties system depends - aside from its explanatory value - on its of a model even in the common case where many parameter consistency with the data available. From a data-driven viewpoint, values are unknown. modeling is an “inverse problem”: given a certain class of models, one tries to identify unknown parameters or even functions which DS2-2-29 give rise to the observed data or a desired qualitative dynamics. In addition to the traditional identification of inaccurately accessible Defining an optimal metabolic operation point based on the rate constants or initial conditions, one can determine the large scale computation of elementary modes topology of a gene regulatory network, or even the functional Terzer, Marco1; Stelling, Joerg2 dependence of a rate law on the species concentrations. In the 1ETH Zurich, Computer Science, Zurich, Switzerland; 2ETH Zurich, presence of data noise, however, parameter identification is an ill- Zurich, Switzerland posed inverse problem in the sense that its solution lacks stability properties. This problem can be overcome by the use of so-called Objective: Organisms have developed control mechanisms regularization methods. to optimize growth under different environmental conditions. Results: One of the systems we study is an ODE model of Optimization techniques such as flux balance analysis have a metabolic pathway, which has been used as a benchmark been used to predict flux values, e.g. maximizing for growth or problem for parameter identification. Our numerical results energy production. However, such single-objective optimization demonstrate the stabilizing effect of regularization on the inverse approaches neglect other essential aspects like robustness problem of identifying all 36 parameters in the 8-dimensional or responsiveness to environmental changes. We show how ODE system. Without regularization, a few percent of data noise to employ elementary modes -- minimal non-decomposable leads to more than 100% relative error in some of the identified pathways -- to define anoptimal metabolic operation point. It is parameters, thus highlighting the ill-posedness of the inverse suboptimal from a single-objective perspective, but optimizes the problem. Using regularization, the relative parameter error is interplay between growth, robustness and responsiveness. We comparable with the data noise. also show how to compute elementary modes efficiently for large Conclusions: The identification of parameters, initial conditions, scale networks. and functions from noisy data is a challenging problem, due to its Results: New algorithmic concepts are introduced enabling ill-posedness. Via numerical examples, we illustrate the nature of elementary mode computation on a large scale. A parallel out- instability in such an inverse problem. A small error in the data can of-core algorithm is developed, with fine-grained parallelization translate to a significant error in the parameters. On a benchmark to exploit multiple cores of modern CPUs, and a coarse-grained problem, we have shown that Tikhonov regularization allows to divide and conquer approach allowing for cluster computations. counter the ill-posedness of parameter identification. In particular, The out-of-core strategy controls memory demand by storing by choosing the regularization parameter appropriately based on intermediary results on disk. the knowledge of the data noise, parameters can be identified in We apply the enhanced computation potential to realistic a stable and accurate manner. metabolic networks. The computed elementary modes (several millions) are used to define an optimal operation point. In addition to maximum yield, we also consider robustness and responsiveness, reflected by a certain degree of flux variability. We analyze the coefficient of variation for all fluxes using modes above different suboptimality thresholds. Interestingly, flux variation increases rapidly when gradually decreasing optimality

ICSB 2008 129 of biomass production to 2% below maximum yield. After this into the vesicle. To analyse spatial dependencies we developed saturation point, we observed only little gain of variation. We call a simulation that tracks the position of relevant molecules and this point optimal metabolic operation point since the cell gains vesicles, reactions as well as vesicle budding and fusion in a most robustness and variability at a small price of suboptimal stochastic framework. biomass production. Results: Simulations show that diffusion alone cannot transport Conclusions: Our methods enable large scale computations of vesicles efficiently from the donor compartment to the target. elementary modes and allow for multi-objective flux analysis. This Despite a low probability for fusion of a vesicle with its donor leads to the definition of an optimal operation point, in our opinion compartment, the rate for such a backward fusion process is a more realistic operation mode than those arising from single- high because the vesicle-compartment distance is not increased objective methods. quickly by diffusion. In contrast, directed movement by motor proteins is only as efficient as the layout of compartment locations DS2-2-30 and the setup of the transport network between them. Conclusion: We have developed a detailed simulation method Modeling mRNA translation using probabilistic model that allows the investigation of the function of sorting and checking transport in endocytosis with respect to parameters such as Bosnacki, Dragan; De Vink, Erik coat or SNARE interactions, concentrations, and transport. Our Eindhoven University of Technology, Eindhoven, Netherlands simulation is able to couple endocytosis with signaling, thus we can include the regulation on signal transduction by endocytosis Objectives: Stochastic models have been widely used to of membrane receptors. The spatial component has a crucial role model biological systems like lambda-phage genetic switches in vesicle transport and must not be neglected. or circadian clocks. In this work we explore probabilistic model checking as a more efficient alternative to stochastic simulations DS2-2-32 for dealing with a broad class of biological systems. The class includes DNA transcription and repair, as well as mRNA Interplay of signal transduction and gene regulation in translation. In these systems probabilities and time play a crucial mammalian cell differentiation role. In particular, as an example of such systems, we focus on Höfer, Thomas Dedicated mRNA translation and we model ribosome kinetics. We calculate German Cancer Research Center, Modeling of Biological Posters the probability for each codon that an erroneous amino acid is Systems, Heidelberg, Germany added to the polypeptide chain as well the average time needed to add an amino acid. In the past decade, major players in T-cell signaling and Methods: We use the model of the ribosome kinetics by interactions between them have been identified. Information Rodnina et al. [3] as developed in [1]. For each codon the model transfer was observed to proceed not only from extracellular divides the tRNAs in three groups: cognate, near-cognate and stimuli ¡V antigen presentation, cytokines, co-stimuli - to the non-cognate, depending on the match between the codon and nucleus, but multiple feedback loops have been found. The the anticodon. We employ probabilistic model checking based consequences of such nonlinear signal processing for the on Markov chains with the model checking tool PRISM[2]. We resulting cellular decisions remain largely unknown. Moreover, check various properties of the systems, like first-passage time given the dynamic nature of the networks, mapping the regulatory probabilities as well as average times. interactions themselves is a complicated task. I will discuss two Results: Our work shows that probabilistic model checking systems on which we employed a joint approach of mathematical and PRISM as a tool in particular are an attractive alternative for modeling and experimentation to get insight into these questions. simulation because the analysis is much more reliable and faster. Studying how regulatory T cells (Treg) control the proliferation of For the translation model we show that the error probability and effector T cells (Teff), we found that competition for the growth the times of each codon are determined by the concentration of factor IL-2 and feedback-enhanced IL-2 signaling are critical aminoacyl tRNAs in the cell. In particular, the error probabilities determinants of the system behavior. The interaction of these are directly proportional with the ratio of the near cognate and processes gives rise to an all-or-none proliferation decision, cognate tRNAs concentrations. efficient modulation of the Teff activation threshold by Tregs, [1] Fluit et al., Computational Biology and Chemistry, 31:335-346, and strong locality of IL-2-mediated cell-to-cell interactions. 2007. With appropriate cytokine stimuli, Teffs differentiate into immune [2] Kwiatkowska et al., Journal of Software Tools for Technology memory cells of various lineages. We have studied the core Transfer, 6:128-142, 2004. network that governs the expression of Tbet (Tbx21), the master [3] Rodnina et al., TRENDS in Biochemical Sciences, 25(2):124- transcription factor of the Th1 lineage. 130, 2004. Iteratively measuring, modeling, and perturbing the transient dynamic behavior of the critical genes, Tbet, IFN-ƒ× and IL- DS2-2-31 12RƒÒ2, we uncovered new regulatory connections. The resulting network consists of interlocked positive feedback Stochastic multi-scale simulation of receptor mediated loops that process the Th1-inducing signals, IFN-ƒ× and IL-12, endocytosis in a defined temporal order. The network dynamics suggest a Klann, Michael; Lapin, Alexei; Reuss, Matthias mechanistic basis for the differential, yet synergistic, effects of the University of Stuttgart and Center Systems Biology, Institute of two inducers on memory cell development. Biochemical Engineering, Stuttgart, Germany DS2-2-33 Objective:The number of membrane receptors is critical for many signaling processes. This number can be regulated by Environmental and regulatory optimization in metabolic endocytosis and subsequent degradation of receptors (e.g. network modeling the pheromone alpha receptor Ste2p in budding yeast), which Byrne, David; Gardner, Timothy; Segre, Daniel involves complex sorting steps in the endosomal system. We Boston University, Program in Bioinformatics, Boston, United assume that the relative localization of the different compartments States in the sorting and degradation system reflect their relationship in the degradation process. Vesicles shuttle between the Objective: Metabolic engineering in microbial hosts for the compartments to sort and transport relevant proteins, and production of renewable chemicals and energy sources has experiments show an overall directed movement instead of received considerable attention in recent years. Microbial fuel undirected diffusion. Protein coded information of the desired cells (MFCs), biohydrogen and biodiesel, as representative transport direction has to be present in a vesicle ab initio. The forms of renewable bioelectricity and biofuel sources, provide efficiency of the whole system in sorting and transport thus new opportunities for the sustainable production of energy from depends on the budding process, where the proteins are packed biodegradable, reduced compounds. Yet MFC electrical current

130 ICSB 2008 and biodiesel fatty-acid synthesis still fall short of the output and DS2-2-35 efficiencies desirable for practical implementation. Since microbial metabolism is the primary cellular mechanism by which energy Genome-scale study of epistasis with respect to multiple is generated and distributed, it is of interest to understand how phenotypes metabolic pathways are regulated, and to develop methods that Snitkin, Evan; Segre, Daniel will increase the proportion of energy diverted for bioenergy use. Boston University, Boston, United States Results: A computational framework has been developed to design nutrient optimization programs that maximize output Objective Epistasis, the phenomenon by which a given gene and efficiency using models of microbial metabolism. Using variant can mask or enhance the phenotypic effect of another the strong duality theorem from linear programming, a bilevel gene, was traditionally observed on a variety of phenotypes, such optimization procedure is used to generate nutrient compositions as organism morphology and color. However, recent studies of that optimally increase bioenergetic yields. To further improve gene-gene interactions in model organisms, such as yeast, have predictive accuracy where regulatory effects are dominant, we largely focused on growth rate and fitness as phenotypes. Here also implement a procedure that uses array-based expression we computationally extend the study of gene-gene interactions profiles to probabilistically infer optimal sets of conditions to a systematic screen of multiple phenotypes, in search for new satisfying logical regulatory rules between input and output interactions, and deeper insights into different cellular functions states. Given one or more transcription regulators of a known or and their interrelationships. inferred target, the method defines the logical on/off state of each Results We used flux balance metabolic modeling in yeast to gene, simultaneously taking account of both target and regulator explore global patterns of epistasis, using as phenotypes the expression levels. predicted steady state rates through all metabolic reactions Conclusions: Optimization predictions indicate that significant (the fluxes).Our analysis involved the simulation of all possible improvements in Shewanella oneidensis MFC electricity single and double gene deletions among the 672 genes present generation and Escherichia coli biodiesel fatty-acid synthesis in the model, and the measurement of ~900 different reaction are attainable. Environment optimization and regulatory network fluxes. The mathematical structure describing all possible reengineering may be used synergistically to increase the potential interactions is a three-dimensional matrix of gene interactions of bioenergetic output. In addition to being applicable to other with respect to multiple phenotypes. The analysis of this matrix metabolic engineering problems, these approaches can help revealed a substantial (at least 5-fold) increase in the number understand the adaptive evolution of metabolic networks and of observed epistatic interactions relative to those observed their regulatory logic. with any single phenotype. Specific interactions relative to non- fitness phenotypes could be validated using available literature,

DS2-2-34 demonstrating the validity of this approach and its potential for Posters future discovery. Dedicated Network-based analysis of longevity and age-related Conclusions Our analysis of epistasis with respect to ~900 diseases using the YABNA platform metabolic phenotypes revealed that growth alone may not be Tacutu, Robi; Budovsky, Arie; Yanai, Hagai; Fraifeld, Vadim sufficiently informative for a comprehensive analysis of function Ben-Gurion University of the Negev, The Shraga Segal Dept of in genetic interaction maps. Rather, many genes may be found Microbiology & Immunology, Beer-Sheva, Israel to interact only upon observing other flux phenotypes. Moreover, the specific phenotype with respect to which two genes Objective: Age-related diseases (ARDs) including atherosclerosis, interact provides information about the functional basis for their cancer, Alzheimer’s disease, and diabetes type II are the major relationship. Our study can help suggest what phenotypes should factor limiting the lifespan in protected human population. Yet, be measured to obtain detailed interaction map for specific the evolutionary and molecular links between ARDs and longevity pathways. remain elusive. We analyzed these links using a network-based approach. DS2-2-36 Results: For this purpose, we have developed YABNA (Yet Another Biological Networks Analyzer), a software program for Nonlinear relaxation phenomena in elastic network models construction, visualization, and analysis of the complex networks. of molecular motors: A comparative study of the functional In particular, for building the human protein-protein interaction motion (PPI) networks analyzed in this study, YABNA uses the data Togashi, Yuichi1; Ueda, Masahiro1; Mikhailov, Alexander S.2; from BioGRID. The program automatically integrates the data Yanagida, Toshio1 on orthologous genes/proteins extracted from the InParanoid 1Osaka University, Graduate School of Frontier Biosciences, Suita, database, and information on the protein function/localization Osaka, Japan; 2Fritz Haber Institute of the Max Planck Society, from HPRD. Comparison of the previously constructed Human Department of Physical Chemistry, Berlin, Germany Longevity Network and the networks of major age-related diseases (Budovsky et al., 2007; 2008) revealed a significant Objective: Elastic network models (ENM; network of material overlap between them. For example, over 85% of nodes and 90% particles and linear springs) are often used to study slow of hubs in the Human Longevity Network are also found in the dynamics of proteins. Directions of conformational changes and Human Cancer Network. Considerable portion of common nodes equilibrium fluctuations in various proteins can be reproduced and especially hubs are also involved in other ARDs and related fairly well by normal mode analysis (NMA) of these models, with processes such as oxidative stress, inflammatory responses, an interesting exception of KIF1A, a kinesin superfamily protein. and wound healing. Apart from the common nodes, there are However, as we pointed out [1], nonlinearity in the relaxation of numerous direct and indirect PPIs between the networks. Using ENMs is significant for large conformational changes often seen in YABNA for analyzing the networks robustness after knock-down biomolecular machines; therefore, relevance of NMA results to the simulations of the selected nodes, revealed potential targets for actual functional motion is not evident. both longevity-promoting and anti-ARDs interventions. In this study, instead of NMA, we directly simulated, compared Conclusions: Collectively, the results obtained suggest that and classified relaxation patterns of ENMs for several molecular common molecular mechanisms stand behind aging and motors, to characterize their motion. ARDs. This supports our hypothesis that ARDs not only are Results: First, we studied relaxation of ENMs starting from direct consequences of aging but actually represent its diverse randomly chosen initial deformations. The models usually show manifestations, being an essential part of the ‘normal’ aging ordered motion along well-defined paths in the conformational process. space. Then, we simulated functional relaxation motion between two machine-states, to study relevance of such ordered paths Supported by the European Union FP7 Health Research Grant to actual conformational changes in the machine cycles. In number HEALTH-F4-2008-202047. KIF1A, the functional motion deviates from the paths starting from random deformations. The relaxation is strongly nonlinear,

ICSB 2008 131 showing a plateau of elastic potential energy. and 11% (9% and 18%), respectively. Conclusions: The elastic network model for KIF1A shows a strongly nonlinear relaxation pattern. A sequence of DS2-2-39 conformational motion (step-by-step reconfiguration) is observed. The model for myosin V shows a much simpler relaxation pattern, Optimized null model for protein structure networks and linear approximation is valid for considerable deformations. Przulj, Natasa1; Milenkovic, Tijana1; Filippis, Ioannis2; Lappe, Such nonlinear behavior of KIF1A may be related to the Michael2 mechanism of the motor [2], which is not well described by NMA. 1University of California, Irvine, Computer Science, Irvine, United We think our method is useful for classification of intramolecular States; 2Max Planck Institute for Molecular Genetics, Berlin, dynamics of biomolecular machines. Germany [1] Y. Togashi, A. S. Mikhailov, Proc. Natl. Acad. Sci. USA 104, 8697 (2007). Objective: In the past few years, the view and understanding of [2] M. Kikkawa, et al., Nature 411, 439 (2001). protein structural space has shifted towards a network (graph) representation of protein structure. In such a framework, protein DS2-2-37 structures are modeled as “residue interaction graphs” (RIGs) where nodes represent amino acid residues and edges describe Universal shape laws on cell and tissue scales pair-wise contacts between residues. Much attention has Bischofs, Ilka1; Klein, Franziska2; Lehnert, Dirk2; Bastmeyer, recently been given to the statistical significance of topological Martin2; Schwarz, Ulrich3 features observed in biological networks. Several studies 1Lawrence Berkeley Lab, Physical Biosciences Division, Berkeley, have related network topological properties, such as network United States; 2University of Karlsruhe, Karlsruhe, Germany; centrality, to protein folding and binding mechanisms, as well as 3University of Heidelberg, Heidelberg, Germany to protein stability and function. Degree-preserving randomized models have been widely used for this purpose in biomolecular Objective: Mechanophysical cues emerge as important decision networks. However, such a single summary global statistic of factors governing cell and tissue differentiation. Geometry can a network may not be detailed enough to capture the complex switch cell fate from proliferation to apoptosis and the shape of topological characteristics of protein structures and their network Dedicated tissue models results in spatially inhomogeneous growth. Here we counterparts. Posters use a model driven approach to study shape determinants across Results: Here, we investigate both local and global network cell and tissue scales. properties of RIGs to find a well fitting network null model for Results: Surprisingly on both scales we find experimentally that them. We compare the network structure of RIGs to several non-adherent edges form circular arcs with an adhesion distance random graph models. The RIGs are derived from a structurally dependent curvature. We explain this finding in the framework diverse protein data set at various distance cut-offs and for of a generalized Laplace law of competing line and surface different groups of interacting atoms. We show that 3-dimensional tension. Model fitting in combination with acto-myosin inhibition geometric random graphs (GEO-3D), that model spatial experiments further reveal that cells control shape by actively relationships between objects, provide the best fit to these RIGs modulating motor tension and contour elasticity. for all reasonable and practically used distance cut-offs. We Conclusions: Our study sets the biophysical foundation for a demonstrate that choosing GEO-3D as a null model results in the quantitative understanding of biological shape determinants that most specific identification of statistically significantly under- and will facilitate rational cell and tissue engineering applications and over-represented subgraphs. suggests that similar strategies might be employed on both scales. Conclusions: To our knowledge, this is the first study that addresses the challenge of finding an optimized null model for DS2-2-38 RIGs, by comparing various RIG definitions against a series of network models. Our findings have important implications for A machine learning approach to infer genetic networks finding statistically significant structural building blocks that play Shieh, Grace Shwu-Rong1; Chuang, Cheng-Long2; Chen, Chung- an important role in protein folding, stability and function. Ming2 1Institute of Statistical Science, Academia Sinica, Taipei, Taiwan; DS2-2-40 2Institute of Biomedical Engineering, National Taiwan University, Taipei, Taiwan Sensitivity analysis based adaptive search-space reduction for parameter estimation applications Motivation: Inferring genetic interactions is of interest since it Liebal, Ulf W.; Schmidt, Henning sheds light on important biochemical pathways. From a group University of Rostock, Rostock, Germany of experiments-confirmed genetic interactions, we observed that paired gene expression curves of transcriptional compensatory Objective: One of the central aspects in systems biology is the interactions often were complementary (anti-similar) whereas estimation of in-vivo kinetic parameters based on experimental those of transcriptional diminished interactions looked similar. This data. Already relatively small models feature a large number motivated us to develop a pattern recognition approach (called of parameters and the usually limited amount of available PARE) to infer genetic networks from time course microarray gene experimental measurement data leads to many of the involved expression data (MGED). parameters being unidentifiable. To improve identifiability, during Methods: PARE learns paired gene expression patterns from the iterative modeling process, new experiments need to be known genetic interactions, either confirmed by biological performed, ideally their design guided by model analysis. Due experiments or from published literature. Specifically, PARE to the usual iteration between data generation and parameter extracts low order characteristics of the nonlinear paired curves, estimation often the question arises which of the parameters are and integrates an optimization algorithm to train the decision the most important ones to be tuned, in order to fit the model to score by MGED of known interactions. Subsequently, PARE can the new experimental data, without destroying the current fit for predict unknown gene interactions of similar nature. the previous experiments. Due to the high dimensionality of the Results: Utilizing yeast MGED in Spellman et al. (1998), PARE search space, often encountered for biochemical models, this predicted 112 pairs of genetic interactions and 77 pairs of question is not answered easily through insight into the system transcriptional interactions (TIs). Checked against qRT-PCR only. results and published literatures, respectively, the modified true Results: We present a new method that allows iteratively positive rates are 73% (70%) and 71% (69%) with n-fold (3-fold) adapting the search space to the most important parameters cross validation, as compared to 52% and 56% of the latest in order to achieve a desired estimation goal. The method is advance in graphical Gaussian models. The false positive rates based on sensitivity analysis of the residuals with respect to the of predicting TC and TD interactions for gene pairs formed from involved parameters. Highly correlated parameters are detected yeast genome (3052 synthetic sick or lethal gene pairs) are 5% and only a less correlated subset is considered further. Employing

132 ICSB 2008 results from linear algebra the method then checks which of key factors in vivo we have developed a minimal in silico model the parameters in the subset are most important to achieve the to investigate the conditions under which the receptor switch desired estimation goal. occurs. Conclusions: The proposed method is a highly valuable addition Results: A differential equation model is studied with respect to to the ensemble of tools and methods that already is available the evolution of HIV in the face of an adaptive immune response. for parameter estimation and modeling tasks in systems biology. The model shows the sudden increase in viral diversity which Although, it is based on linear approximation it has been shown goes along with the impairment of the immune response as can to provide modelers with crucially needed information, allowing be observed at the onset of AIDS. In this framework we can for a faster workflow in modeling. The method can be used both study minimal conditions that lead to the switch in coreceptor manually by the modeler and be build into parameter estimation usage. Conditions that lead to a coreceptor switch in silico tools to be used automatically. Along with the theoretical such as variable immune responses and changes in the target description of the method we provide an implementation of the cell populations help to strengthen hypotheses regarding the method in the Systems Biology Toolbox 2 for MATLAB (www. dynamics in vivo. sbtoolbox2.org). Conclusions: A minimal model of HIV evolution allows to investigate key factors that drive the switch of coreceptor usage DS2-2-41 from CCR5 to CXCR4 such as variations in the immune response or in the target cell populations. Discovering geometric structure in protein-protein interaction networks: The embedding algorithm DS2-2-43 Przulj, Natasa1; Higham, Desmond, J.2; Rasajski, Marija1 1University of California, Irvine, Computer Science, Irvine, United From stationary and nonstationary isotope labeling States; 2University of Strathclyde, Department of Mathematics, patterns to metabolic fluxes: A comparative study for a Glasgow, United Kingdom lysine producing strain of Corynebacterium glutamicum Nöh, Katharina1; Noack, Stephan1; Wiechert, Wolfgang2 Objective: Networks (also called graphs) are widely used to 1Research Centre Jülich, Institute of Biotechnology, Jülich, describe natural phenomena explored in computational and Germany; 2University of Siegen, Siegen, Germany systems biology. Understanding, summarizing and extracting information from these networks are key goals in systems Objective Intracellular metabolic flux rates are the manifestation biology. Constructing a good network null model for protein- of the metabolic pathway activity in organisms. Metabolic Flux protein interaction (PPI) networks is a fundamental issue, since Analysis (MFA) with isotope labeling experiments (LE) enables such a model would provide insight into the biological structure to effectively resolve in vivo not directly measurable flux rates Posters and function and could be used as a guide for further biological and, thus, has become a key technology in Systems Biology. Dedicated experiments. It has been established that geometric random The classical, stationary isotope MFA characterizes a cell’s graphs give an excellent fit for various global and local measures fluxome in a metabolic and isotopic stationary state. Typically, of PPI network structure such as pathlengths, clustering in a continuous culture the feed is switched from naturally to coefficients, relative graphlet frequencies, and graphlet degree isotopically labeled medium which propagates though the distributions. network and progressively replaces the unlabeled intermediates. Results: We design an algorithm to test directly whether When the labeling patterns are approximately time-invariant, PPI networks are geometric by embedding them into a low samples are taken. Isotopically non-stationary MFA is a novel area dimensional Euclidean space. If a geometric network model fits of research which represents a promising generalization of the the PPI data, then it is expected that PPI networks could be classical approach. Cells are likewise kept under metabolic steady embedded into the Euclidean space. The algorithm is based on state conditions, however, now the time profiles of the labeling Multi-Dimensional Scaling, with pathlengths playing the role of patterns are measured upon start of the labeling period to monitor the Euclidean distances. We judge the sensitivity and specificity the labeling propagation. of the fit by computing the areas under the Receiver Operator Results C. glutamicum is a promising candidate for industrial Characteristic (ROC) curve. We test the algorithm on 19 PPI lysine production. To understand the intracellular flux distribution networks of four different organisms, as well as on artificial to lysine and alternative products a fed-batch cultivation is carried networks corresponding to a high-confidence yeast PPI network out. The specifically labeled glucose mixture is determined in an that were generated using 7 types of random graph models. The a priori experimental design study. Twenty samples are taken in resulting areas under the ROC curve (AUCs) when real-world PPI sub-second scale following an exponential time profile until an networks and geometric random graphs are embedded into 2D, isotopic pseudo-steady state is reached. LC-MS/MS is applied 3D, or 4D space, are significantly higher than AUCs obtained to simultaneously analyze labeling enrichments and pool sizes when embedding any of the non-geometric random graphs. for intermediate metabolites as well as amino acids. Together Conclusions: We provide a method for embedding networks into with glucose consumption and product secretion rates detailed low dimensional Euclidean space. The method provides a direct intracellular flux maps for both approaches are determined, test of whether PPI networks have a geometric graph structure. compared and discussed. By linking between fluxome and Our results yield support to the hypothesis that the structure of metabolome data a cross-check becomes possible. currently available PPI networks is consistent with the structure of Conclusion Compared to the classical isotope MFA, non- geometric graphs. stationary LEs are not only much less time-consuming (minutes versus hours), but also cheaper and desirable due to the high DS2-2-42 information content of the measurement data. Moreover, the newer approach is closer to production conditions predominating A minimal model for the HIV coreceptor switch industrial applications and imperative for higher cells. Kamp, Christel Paul-Ehrlich-Institut, Biostatistics, Langen, Germany DS2-2-45

Objective: The entry of HIV into its target cells is facilitated by the Quantification and analysis of Aspergillus niger metabolic prior binding to the cell surface molecule CD4 and a secondary network via nullspace approach coreceptor, mostly the chemokine receptors CCR5 or CXCR4. Melzer, Guido; Eslahpazir, Manely; Nörtemann, Bernd; Hempel, In early infection CCR5-using viruses are mostly dominant while Dietmar Christian; Franco-Lara, Ezequiel; Melzer, Guido a receptor switch towards CXCR4 occurs in about 50% of the Technische Universität Braunschweig, Institute of Biochemical infected individuals which is associated with a progression of the Engineering, Braunschweig, Germany disease. There are many hypotheses regarding the details of the complex interplay between HIV and the immune response that Objective: Aspergillus niger is one of the most important try to explain the coreceptor switch. While it is difficult to isolate filamentous fungi for the production of a wide variety of organic

ICSB 2008 133 acids and relevant biocatalysts. Knowledge about the minimal the fault tolerance abilities of our new model, which we believe will sets of enzymes that are required for the efficient production be higher than that of classical RBNs. of a desired protein under steady state cultivation conditions could substantially contribute to the design and realization of DS2-2-47 an optimized production process. Therefore, metabolic network analysis using elementary modes was used to quantify the NAD(P)H waves and calcium oscillations in neutrophils ? A biochemical reaction system of A. niger for protein production. spatiotemporal modeling study of feasibility Results: Pathway analysis by nullspace approach [Wagner Lebiedz, Dirk; Slaby, Oliver 2004] was applied and implemented in MatLab as a powerful University of Freiburg, ZBSA (FRISYS), Freiburg, Germany mathematical tool to estimate elementary modes, which provide concrete insights into the crucial reaction pathways generating Exotic metabolic wave phenomena together with mutually phase- only products of interest under distinct experimental conditions. coupled NAD(P)H and calcium oscillations have been observed However, the formation and excretion of large amounts of in human neutrophils. At least parts of these phenomena are oxalic acid at external pH-values of ≥ 5 represents an undesired doubtful due to extensive failure of reproducibility by several misrouting of the carbon source into a valueless side product. other groups. The aim of our theoretical spatiotemporal modeling This loss, for instance, can account for more than 30% of the total approach is to propose a possible and plausible biochemical carbon content of the respective carbon source (e.g., D-xylose, mechanism which would in principle be able to explain such D-glucose) under process conditions at pH 5.5 and is therefore oscillation and wave phenomena in neutrophils. Our modeling counterproductive for the production of a target protein. suggests the possibility of a calcium-controlled glucose influx as Since the consumption rates of the substrates (e.g. glucose, a driving force of metabolic oscillations and a potential role of ammonium) as well as the production rates (e.g., for target polarized cell geometry and differential enzyme distribution for protein, biomass, oxalic acid) depend on the variation of process various NAD(P)H wave phenomena. The modeling results are conditions; the actual composition of the cultivation medium supposed to stimulate further controversial discussions of the provides an indication about the active pathways and revealed phenomena and experimental efforts to finally clarify the existence a maximal protein production yield coupled by high yield for and biochemical basis of any kind of temporal and spatiotemporal oxidative pentose-pathways for supplying sufficient NADPH. patterns of calcium signals and metabolic dynamics in human Dedicated Conclusions: The analysis of the active elementary modes for neutrophils. Independent of published observations, they present Posters production of target protein provided crucial insights into the a general feasibility study of such phenomena in cells. functionality of the A. niger metabolic network. The calculation of optimal yields for desired products at minimal yields for the DS2-2-48 undesired by-products can enable the development of potential knock-out strategies and thus lead to an optimized (and Automatic complexity analysis and model reduction of economic) production process. nonlinear biochemical systems Wagner, C. (2004) J. Phys. Chem. B 108, 2425-C2431. Fein, Marc; Skanda, Dominik; Lebiedz, Dirk University of Freiburg, Center for Systems Biology, Freiburg, DS2-2-46 Germany

How spatial and temporal choices influence the dynamics Objective: Kinetic models for biochemical systems often of generalized boolean networks comprise a vast amount of coupled differential equations with Darabos, Christian1; Giacobini, Mario2; Tomassini, Marco1 species concentrations varying on different time scales. We 1University of Lausanne, Information Systems Institute (ISI-HEC), present and apply two new and efficient methods for automatic Lausanne, Switzerland; 2University of Torino, Computational complexity analysis and model reduction. Biology Unit, MBC, Torino, Italy Results: A combination of these two methods have successfully been applied to a biochemical example model describing the Objective: Over thirty years have passed since Kauffman’s regulatory influence of RKIP on the ERK signaling pathway. original Random Boolean Network model, focusing on the We numerically compute a reduced model capturing the active system-level properties of Genetic Regulatory Networks as system dynamics by exploiting the results of the complexity a whole. Since then, our understanding of the structure and analysis. dynamics of biological systems has increased tremendously. Conclusions: The first method combines dynamic sensitivity Therefore, some of the early assumptions are now known to be analysis with singular value decomposition to determine the over-simplistic. Originally, RBNs rest on three main hypotheses: minimal dimension of the kinetic model necessary to describe the the nodes’ values and functions are Boolean, the network active dynamics of the system accurately enough within a user- topology is random and the system is synchronous. We propose defined error tolerance for particular species concentrations and changes that, in our opinion, bring the model closer to current to determine each species’ contribution to the active dynamics. known facts about genetic regulatory networks. Firstly, present The second method treats the explicit numerical reduction of the data suggest that biological networks are of the scale-free model to a lower dimension according to the results of the first or hierarchical type, but not random, we call it Generalized method and allows any species combination to be chosen as Boolean Networks (GBNs). Secondly, many recent experimental a parameterization of the reduced model which may either be observations tend to prove that genes are expressed according tabulated in the form of look-up tables or computed in situ during to a strict sequence. Thus we propose a new semi-synchronous numerical simulations. A reduced representation of a multiple update sequence where we consider the influence of one node time scale system is particularly advantageous in the context on another as an active biological expressing or repressing of spatiotemporal simulations which require high computational factors. We have called this update scheme the Activated efforts. The benefit of combining the two methods lies in the Cascade Update (ACU). fact that a user, who is interested in model reduction, can Results: We compare the effect of standard Synchronous choose the dimension of the reduced model and the reaction update and our ACU scheme on both classical RBNs and GBNs progress variables parameterizing the reduced model freely in with scale-free topology of different sizes in terms of number of accordance with the problem-determined needs on the basis of attractors, and their length. We show that the ACU steadily leads the complexity analysis. to significantly more attractors than Synchronous update. We also show that scale-free networks find in general more unique attractors of short length, which are the only ones believed to have biological significance. Conclusions: We have successfully developed a model with characteristics closer to those of real biological networks, which also show better dynamical behavior. Our current research studies

134 ICSB 2008 DS2-2-49 biomolecules and putative target genes and information about the distribution of target genes, a gene regulatory network was Evolution of asymmetric division: A modelling approach developed. Rashidi, Armin; Shanley, Daryl Conclusions: The cluster analysis was a good method to detect Institute for Ageing and Health, Newcastle University, Newcastle genes regulated similarly in the development of insulin resistance. Upon Tyne, United Kingdom The gene regulatory network explained the distribution of genes in the three mouse groups and confirmed the regulation of several Objective: Symmetric reproduction precludes ageing; all known target genes. individuals would be affected by any deterioration and the lineage would vanish. The recent discovery of asymmetric division in DS2-2-51 bacteria suggests that ageing evolved early in the history of life. Segregation of damaged macromolecules to one progeny cell OpenMicroBio: A framework for the simulation of cellular results in an ageing parent and a rejuvenated daughter. This communities during bioprocess engineering observation shifts the question of why ageing evolved to that Costa Martins, Rui1; Castro, Cristiana2; Vicente, António2; Teixeira, of why asymmetry evolved. Using in-silico experiments, we José2; Rosa, Agostinho3; Lopes, Vitor3 here show that selection pressure promotes the evolution of 1Molecular and Environmental Biology Research Center, University asymmetry under certain circumstances. The assumptions used of Minho, Braga, Portugal; 2IBB - Institute for BioTechnology and in the model are: (i) Limited resource availability creates a trade- BioEngineering, Braga, Portugal; 3ISR - Institute for Systems and off between investment in reproduction and in maintenance, (ii) Robotics, Lisbon, Portugal The population is near its carrying capacity and the optimum investment strategy has already evolved, (iii) The rate of damage Objectives: OpenMicrobio is an ‘open source’ on-going project accumulation (r) is inversely related to the level of investment in to build a framework for the simulation of cellular communities maintenance, (iv) The interval to next division (Δt) is shortened during bioprocess engineering, describing the Saccharomyces by larger reproductive investments, and (v) The likelihood of colony dynamics based on individual cell models (ICM), using survival up to a certain time decreases with the time-integrated complex systems approaches. The framework architecture is damage up to that time. The damage segregation coefficient,σ , devoted for enabling multiscale simulations from the microscale to was allowed to evolve between zero (full symmetry) and one (full the macroscale levels, by simulating both colony dynamics under asymmetry), and was averaged over the population at any given several scenarios, such as, production inside bio-reactors and time. The outcome of each run was defined as a( ) asymmetry: σ > growth in fermented foods, where the colony is highly affected 0.8 (more than 90% of damage segregating to one progeny cell) by external such as as fluid dynamics, chemical and biochemical for at least 50% of the monitoring period or (b) symmetry: σ > 0.8 reactions, nutrient diffusion and electromagnetic fields, such as Posters for less than 25% of the monitoring period. during flocculation, stress and morphological adaptations to Dedicated Results: Two stochastic parameters (random mutations and metabolites and communication molecules (e.g. pheromones and the likelihood of survival to next division) and three fundamental aromatic alcohols). constants (Δt, C1, C2) are potential determinants of the dynamics Results: We herein present the modelling strategy and system of the system. C1 determines the shape of the relationship architecture under development. The framework is devoted to between r and Δt, and C2 is the level of time-integrated damage build the systems libraries that will enable to build-up simulation above which the chance of survival is zero. Sufficiently (but not scenarios from different information levels: i) individual yeast cell too) large Δt, sufficiently convex relationship betweenr and Δt, cycle and morphology changes by developing the ICM models and sufficiently small C2 are factors that promote the evolution of using artificial life techniques (cellular automata paradigms); ii) asymmetric division. characterizing cell-to-cell interactions by passive (e.g. nutrient Conclusions: We determined, by a model, the conditions or metabolite diffusion) and active (pheromones during mating) required for the evolution of asymmetric division. communication using physical biology principles such as electrochemical-diffusion and microfluidics; iii) build-up the colony DS2-2-50 overall behavior, leading to quorum sensing behaviors based on communication mechanisms; and iv) integration with fluid A gene regulatory network of insulin resistance in mice dynamics, chemical/biochemical reactions and nutrient diffusion Schoettler, Anja1; Pospisil, Heike2; Scheja, Ludger1; Beisiegel, (Lattice-Boltzmann diffusion, micro-fluidics and particle collision/ Ulrike1 electromagnetic interaction). 1University Medical Center Hamburg-Eppendorf, Biochemistry Conclusions:The Openmicrobio architecture uses continuous and Molecular Biology II, Hamburg, Germany; 2University of approaches (e.g. force fields, diffusion) with discontinuous particle Hamburg, Center for Bioinformatics, Hamburg, Germany mechanics and Cellular Automata, which allows to control different phenomena and simulation scale, being a promissing Objective: Several genes are involved in the development of modelling approach for building user-friendly BioEngineering insulin resistance. In previous studies, the regulation of these simulation systems. genes was mainly studied by activation or inactivation of a single transcription factor. Our aim was to develop a gene regulatory DS2-2-52 network which takes into consideration the interplay between different transcription factors in the murine liver as well as different Fragile, yet persistent: The cell as autonomous self- stages in the development of insulin resistance. fabricator Results: Gene expression (TaqMan real-time PCR) and several Hofmeyr, Jan-Hendrik S.1; Wolkenhauer, Olaf2 biomolecules relevant for regulation of transcription factors were 1University of Stellenbosch, Biochemistry, Stellenbosch, measured in three groups of C57Bl/6 mice displaying different South Africa; 2University of Rostock, Systems Biology and degrees of insulin resistance. Gene expression data of all groups Bioinformatics, Rostock, Germany were combined and Spearman Rank correlations were computed between all genes. The resulting correlation coefficients were Objective: One of the most, if not the, distinguishing features subsequently used in a cluster analysis which detected three of living systems as we know them is the ability to persist as main gene clusters. Correlations between genes and several a functional entity despite the fragility of all of its components: biomolecules proved that gene clusters were regulated by distinct a living system survives the lifetimes of all its components. insulin-regulated transcription factors, presumably SREBP1c, This implies that living systems are able to autonomously self- SREBP2, ChREBP, LXRα, Foxo1 and Foxa2. PPARα target genes fabricate themselves. The objective of this study was to develop were found in each of the clusters, assuming a major role for a formal model of the functional organisation required to make insulin-regulated transcription factors compared to PPARα in the self-fabrication possible. It builds on, reconciles and extends development of insulin resistance. the groundbreaking work of Robert Rosen (M,R-systems and Combining information from the literature, correlations between relational biology), John von Neumann (self-reproducint automata)

ICSB 2008 135 and Marcello Barbieri (semantic biology). A preliminary account DS2-2-54 can be found in [1]. Results: The formalisation of Aristotelean explanatory factors Mathematical modeling of cellular self-replication from a (causes) in terms of category theory allows the development physicochemical perspective of a formal model that captures the causal structure of a self- Lindahl, Paul1; Surovtsev, Ivan1; Morgan, Jeffrey2 fabricating system. A novel treatment of formal cause allows 1Texas A&M University, Chemistry, College Station, Texas, United the resolution of a purported paradox noted by Rosen in Von States; 2University of Houston, Mathematics, Houston, Texas, Neumann’s description of kinematic self-reproducing automata. United States It is shown that supramolecular chemistry (self-assembly) is the key to closing the loop of efficicient causation that makes self- Objective: The objective of this NSF-sponsored project is to fabrication, and therefore life, possible. understand how cellular self-replication might occur on the Conclusions: The study of functional organisation requires a new mechanistic biochemical level. We have developed a modeling formalism: category theory provides a natural language for this framework that assumes a cell consisting of cytosol and purpose. Previously [2] we showed that the traditional formalisms membrane. Cells grow according to user-defined mechanisms used in systems biology, such as systems of ordinary differential in which all cellular components are derived from environmental equations, cannot capture the causal structure of self-organising nutrients. Including interfacial reactions renders concentration systems, whereas the theory of cartesian closed categories can. dynamics sensitive to cell geometry. Mechanisms and parameters Here we show how to treat self-fabrication within this formalism. are adjusted to achieve self-replicative behavior, defined as when [1] Hofmeyr, J.-H.S. (2007) The biochemical factory that the surface/volume ratio changes periodically and when both autonomously fabricates itself: A systems-biological view of surface and volume double during that period. Our initial objective the living cell. In Systems Biology: Philosophical foundations (F. was to construct the minimal in silico cell model that exhibited Bruggeman, F.C. Boogerd, J.-H.S. Hofmeyr, H.V. Westerhoff, these properties. Subsequent objectives involved replacing eds.) Elsevier, Amsterdam, pp. 217-242. minimal “modules” with more realistic counterparts. [2] Wolkenhauer, O. and Hofmeyr, J.-H.S. (2007) An abstract cell Results: The minimal in silico cell model that exhibited self- model that describes the self-organization of cell function in living replicative behavior included five symbolic components and systems, J. theor. Biol., 246, 461-476. reactions. A second generation cell was developed by replacing Dedicated the “cell cycle engine” in the minimal model with a more realistic Posters DS2-2-53 mechanism of mitotic regulation. Time-dependent changes in surface and volume restrict possible cellular geometries, as do Reverse engineering algorithms for genome-wide membrane bending energies. These considerations suggested regulatory networks have different performances on stable that our in-silico cell models do not “pinch” at midcell during and causal interactions division, in contrast to real cells. To evaluate whether the cellular Zampieri, Mattia; Soranzo, Nicola; Altafini, Claudio cytoskeleton might enforce pinching, a chemical model describing SISSA/ISAS, International School for Advanced Studies, Trieste, the assembly, steady-state dynamics, and contraction of the FtsZ Italy ring, as found in prokaryotic cells, was developed. This module is currently being integrated into our most complex whole-cell Objective: The goal of a genome-wide reverse engineering model, hopefully to enforce pinching behavior. method is to infer putative gene-gene interactions from the Conclusions: A framework based on standard physicochemical analysis of compendia of microarray data. Several studies laws has been developed for modeling growth and division of suggest that gene co-expression is mainly associated to stable whole cells. A number of cells has been designed, the most relationships, like belonging to the same protein complex, recent of which has focused on the role of the cytoskeleton in although other interactions more of a “causal” and transient cell division. Future studies will involve modeling the actomyosin nature (e.g. transcription factor-binding site interactions) are also ring used in animal cell cytokinesis. A post-doctoral position is possible. The aim of this work is to verify if network inference available. algorithms are indeed useful in discerning stable from causal connections in both artificial and real gene networks. DS2-2-56 Results: We consider five similarity measures: two direct (Pearson correlation and mutual information) and three conditional The expected time paradox of a bistable switch (partial Pearson correlation, conditional mutual information Steijaert, Marvin; Hilbers, Peter; ten Eikelder, Huub and graphical Gaussian model). The first two metrics are Eindhoven University of Technology, Biomedical Engineering, based on gene-gene co-expression, while the remaining three Eindhoven, Netherlands perform a conditioning operation on the two-point measure. We apply these algorithms to an artificial dataset (generated Most stochastic biochemical systems are too complicated to be by a system of non-linear ODEs associated to a network with suitable for an analytical approach. In those cases, simulation a characteristic topology) and two microarray collections (for algorithms such as Gillespie’s SSA are required. E.coli and S.cerevisiae). The reconstructed networks are then In contrast, we focus on systems that are simple enough to allow compared to the true ones; for the real data, we use both known an analytical approach. We have studied a single phosphorylation protein complex networks and transcription factor-binding site cycle with trans auto-phosphorylation, using a parameter set that interactions collected from literature. The comparisons show that yields bistability. One remarkable result is the apparent divergence the gene interactions associated to protein complexes are better in expected times between analytical and simulation results. detected by the direct methods, while those associated to the Consider the equilibrium distribution over all possible states, effect of (multiple) transcription factors are better retrieved by the ranging from a state with all proteins de-phosphorylated to a conditional metrics. state with all proteins phosphorylated. For the bistable parameter Conclusions: Even in the regime of weak inference power we set, this distribution has two local maxima (one of which being have to work in, our analysis confirms that the performances the global maximum of the distribution) and an intermediary of the algorithms are coherent with the features they are meant local minimum. The state space of such a system contains to extrapolate from the data (direct for stable interactions, some regions in which the (theoretical) expected time to get one conditional for causal relationships). reaction step closer to the local maximum grows exponentially with the total number of particles N. For small N, the expected time found with a reasonably large number of simulations is in good agreement with the theoretical time. However, when N is large enough, the simulation results will eventually decrease with N. The explanation for this paradoxical behavior is the existence of rare long paths in the opposite direction (i.e., away from the

136 ICSB 2008 local maximum). Although the probabilities of those rare paths parameters and network topologies into a prior distribution. While decrease with N, the corresponding times grow even faster with this prior could be used to include arbitrary additional information, N, yielding an increasing influence on the expected time. As a we discuss particularly choices for sparse and scale-free result, the amount of simulations that is required for sufficient topologies. sampling of this behavior grows exponentially with N as well. Results: We have tested our approach on simulated data and on Conversely, as most biological interesting processes take real data on the yeast cell cycle. We demonstrate the feasibility place on relatively small timescales, our work indicates that the of the approach on a simulated 3 gene network, showing that expected time is not a suitable variable for the analysis of such dynamic behavior (oscillations) and network parameters can systems. be adequately reconstructed. The effect of different levels of noise and number of data points is studied. We then show the DS2-2-59 applicability also to larger networks with up to 10 genes. Results on real data show that key interactions can be recovered, but A mathematical model for a neuronal polarization also point to limitations of microarray gene expression data for Naoki, Honda quantitative modeling in light of posttranscriptional events, high Kyushu University, Department of Biology, Faculty of Sciences, noise, and low measurement numbers. Fukuoka, Japan Conclusions: We have developed a novel approach to infer networks from time series data, by combining delay differential Objective: A neuron has a polarized morphology consisting of a equations with a Bayesian approach. While examples we show soma, an axon and dendrites, in order to transfer received electric are for gene regulatory networks, the approach can also be signals to the next neurons in one direction. This polarization is extended to signal transduction networks. By sampling from established during development. At the beginning, a cell produces the posterior distribution, we can derive probability distributions several neurites with similar lengths. Subsequently, abruptly only over network topologies and paramters, which holds interesting one of them begins to rapidly elongate; this neurite grows to an promises for experiment design. Work on this is presently axon, and the others to dendrites. Even if a neuron is cultured in a ongoing. uniform chemical condition, it can still be polarized, indicating that developing neurons have an ability to spontaneously break their DS2-2-61 own morphological symmetry. In the polarization process, various kinds of molecules involved in the neurite extension are actively Stochastic modeling of single cell dynamics using discrete- and stochastically transported by motor proteins from the soma time Markov chains to the neurite tips. In this study, we mathematically examined Samaga, Daniel; Kremling, Andreas the mechanism by which a neuron spontaneously polarizes, Max Planck Institute for Dynamics of Complex Technical Systems, Posters based on both the intrinsic stochasticity and the effects of its own Magdeburg, Germany Dedicated morphology. Results: First, we constructed a biophysical model of intracellular Objective: Nonlinearities in cellular systems are known to be reaction networks in a neurite tip. A bifurcation analysis showed responsible for the realization of distinct functionalities. Since that as the concentration of the transported molecule (PI3K- some of them such as hysteresis of the lac operon in Escherichia activating factor) reaches a certain threshold at a neurite tip, coli act inherent stochastic, a shift of bifurcation points can be the neurite elongates to be an axon ( eaxon determination f). expected (Kepler, Biophys. J., 2001). Most dynamic models in This phenomenon was found to be spontaneously induced by this field summarize molecular mechanisms in kinetics and hence stochastic transportation. Next, we examined how the growth of start at a high level of idealization. To keep in view all possible other neurites was suppressed after the axon determination. Our sources of stochasticity and to avoid uncertain assumptions, a mathematical analysis revealed the following logic of the neuronal model based on biochemical reactions of the mass action type polarization. If a certain neurite elongates, the ratio of PI3K- is to set up. The model then retains high level of detail in the activating factors returning to the soma by diffusion is significantly molecular mechanisms by the price of few more parameters. decreased because of the degradation of the molecules. Advantageously, this set of reactions delivers access to stochastic Then, the concentration of PI3K-activating factors in the soma simulation via Gillespie’s exact algorithm (SSA). Because SSA decreases, and the amount of their transportation decreases simulations are very slow in stiff models, approximating algorithms as well. Consequently, the probability that the concentrations of are needed. Therefore a method that uses SSA efficiently will be PI3K-activating factors at the tips of the remaining neurites reach introduced, in which the use of elementary reactions keeps the the bifurcation threshold significantly decreases. level of transparency high. Conclusions: We presented a mathematical framework to Results: An ODE model (Mettetal, PNAS, 2006) was transformed understand how developing neurons are spontaneously polarized. into a set of nine elementary reactions that can be written in terms of the chemical master equation (CME). In deterministic DS2-2-60 simulations the CME behaves just like the ODE. However, quantitative and qualitative differences occur between the Statistical inference of biochemical networks with delay stochastic and deterministic simulations of the CME. For differential equations approximation of the single cell dynamics we find in preliminary Kaderali, Lars; Mazur, Johanna; Ritter, Daniel results Markov chains sufficient. Discrete-time Markov chains are University of Heidelberg, ViroQuant Research Group Modeling represented as transition matrices. Thus they only need a number (BQ26), Heidelberg, Germany of states and transition probabilities among them. Now each SSA simulation contributes to an entry of the transition matrix and Objective: The inference of signal transduction and gene computational effort can be distributed efficiently. Once the matrix regulatory networks (GRN) from data is a major challenge in is calculated, dynamics of freely chosen distributions of molecule systems biology. We present a novel approach to infer GRN numbers are predicted in seconds. from time series gene expression data, using delay-differential Conclusions: Kepler’s theoretical results are shown to be equations to describe the system’s dynamic behaviour. relevant for the lac operon. The range of concentrations providing Parameters of the differential equations and network topology are bistability is enlarged by 25% in the stochastic setting. We estimated simultaneously using simulated annealing or MCMC, introduce an efficient procedure that ends in a very fast method of the latter permitting the computation of distributions over network predicting single cell gene expression dynamics. topologies and model parameters. The number of free parameters in such a model usually far exceeds the number of datapoints available. This leads to underdetermined optimization problems, and overfitting. To circumvent this, we embed our differential equations into a Bayesian framework, and encode expectations on model

ICSB 2008 137 DS2-2-62 mechanism that is ordered and symmetric, and that association, and dissociation, of the proton is the step that limits the turnover- Multiple dynamic characteristics of incoherent feed- rate of MCT1. The model suggests that CAII increases the forward loops effective rate constants of the proton reactions, possibly by Kim, Dongsan1; Kwon, Yung-Keun2; Cho, Kwang-Hyun1 working as a proton antenna. 1Korea Advanced Institute of Science and Technology (KAIST), Conclusions: The dependencies and the particular form of the Department of Bio and Brain Engineering, Daejeon, Republic of rate expressions obtained by the model reduction served as Korea; 2University of Ulsan, School of Computer Science and guidelines in designing experiments that contained the necessary Information Technol, Ulsan, Republic of Korea information to discriminate between models. Furthermore, it was insights gained from the modeling and the model reduction that Objective: An incoherent feed-forward loop (FFL) is one of led to an extension of the MCT1-model to also incorporate the the most frequently observed motifs in signal transduction effect of CAII. This illustrates that modeling itself is a valuable pathways and genetic regulatory networks. This is intriguing if source of new ideas and hypothesis. we consider its structure which seems very inefficient at first glance. A particular dynamics induced by such an incoherent DS2-2-64 FFL is the biphasic behavior which can be classified into two types: time- and dose-dependent biphasic responses. We note Defining the dynamical backbone of gene regulatory that many observed incoherent FFLs have either only dose- or networks time- dependent biphasic dynamics. But, sometimes they show Rodriguez-Caso, Carlos; Corominas-Murtra, Bernat both time- and dose-dependent biphasic dynamics although Universitat Pompeu Fabra (PRBB-GRIB), Complex Systems Lab, their structures are mostly same. To investigate the dynamical Barcelona, Spain characteristics of incoherent FFLs, we have constructed a simple model and analyzed its dynamic behavior. Objective: The inclusion and merging of dynamics into Results: Throughout extensive computer simulations, we topology constitutes the next challenge for uncovering the found that the parameter range showing the time- or dose- dynamical organization of gene regulatory networks. The dependent biphasic response is quite limited. Moreover, the genomic scale approaches have provided a confident picture of Dedicated optimal parameter set leading to a maximal time-dependent gene transcriptional regulatory networks (GTRN) for yeast and Posters biphasic response is far different from that of a maximal dose- E. coli models. Topological approaches have revealed a non dependent biphasic response. More interestingly, we found that trivial organization that strongly departs from simple random an incoherent FFL evolved to have a time-dependent biphasic homogeneous metaphors. However, the resulting picture is response and a dose-dependent biphasic response contain essentially static and poorly captures the topological traits leading different dynamics. to dynamical constraints. The consideration of the directed nature Conclusion: Through the previous results, we found that the of the graph as a causality relation between genes can help us in incoherent FFL can generate multiple dynamic and functional overcoming this limitation. characteristics. Actually, it is known that the structural feature of Results: Under this perspective, we have identified the minimal the network does not always guarantee the functional dynamics. fraction of genes and their relations -the dynamical backbone So, we suggest that various incoherent FFLs might have been (DB)- responsible for the network dynamics. DB is a parameter designed to achieve their specific functional roles under different free definition where the resulting graph emerges from the cellular contexts. application of an extremely simple rule: the iterative pruning of vertices without outdegree. The hierarchical character of the DS2-2-63 GTRNs is revealed by collapsing the cyclic pathways into a single vertex. These cyclic pathways can be considered as dynamical A model reduction approach to the kinetics of the modules since they constitute irreducible units of computation. monocarboxylate transporter MCT1 and carbonic The analysis of yeast and E. coli GTRNs revealed comparable anhydrase II sizes of DB although they differ in their internal organization. Almquist, Joachim1; Schmidt, Henning2; Lang, Patrick3; Prätzel- Interestingly, E. coli DB presents a strong hierarchical structure Wolters, Dieter4; Deitmer, Joachim4; Jirstrand, Mats1; Becker, compared with its yeast counterpart. Our results revealed a large Holger4 and single dynamical module covering half of the yeast DB in 1Fraunhofer-Chalmers Centre, Göteborg, Sweden; 2Universität contrast with the small dynamical modules observed in E. coli. Rostock, Rostock, Germany; 3Fraunhofer Institut für Techno- und The fundamental differences observed in the internal organization Wirtschaftsmathematik, Kaiserslautern, Germany; 4University of of gene regulatory networks in these organisms provide essential Kaiserslautern, Kaiserslautern, Germany clues for the understanding of their complexity. Conclusion: We propose DB and dynamical module as Objective: According to the astrocyte-neuron lactate shuttle topological-based definitions that yield relevant clues about how hypothesis, monocarboxylate transporters (MCTs) play a crucial topology pervades dynamics. role in brain energy metabolism. The monocarboxylate transporter isoform I (MCT1) facilitates transport of lactate and other energetic DS2-2-65 compounds together with protons across the glial cell membrane. As also reported for other acid/base transporting proteins, the Bioψ: A formal modelling language to describe biological transport activity of MCT1 is increased by the enzyme carbonic processes using elementary brick of action anhydrase isoform II (CAII). To describe the transport kinetics Peres, Sabine; Molina, Franck of MCT1, and the impact of CAII, we apply a combination of CNRS, Montpellier, France electrophysiological techniques and mathematical tools such as modeling using ordinary differential equations and model Objective: There is an obvious need of formal description reduction. of biological processes; current databases naturally describe Results: Model reduction techniques aim at simplifying models processes from molecular and cellular point of view but they to reach an appropriate level of detail for experimental validation. rarely take into account the multi-functionality of a molecule, the We have explored a range of model reduction assumptions context and they do not dissociate the function to the molecule based on substrate binding order and timescale separation. itself. Therefore, processes must be described with new angle Each assumption resulted in a unique, explicit transport rate of view. The set of processes is large and heterogeneous and expression, constituting a model for the substrate flux of most of them can be decomposed in combination of more simple MCT1. Simulations of the different models were compared with processes. experimental data obtained from MCT1-expressing Xenopus Results: We propose a computational approach, Bioψ, allowing oocytes injected with different amounts of CAII. Based on the modelling of biological network at different levels of scale single substrate inhibition experiments we propose a binding taking into account the biological context based on elementary

138 ICSB 2008 brick of action and allows the description of multi-level biological DS2-2-67 processes from sub-molecular details to network. Moreover, Bioψ dissociates the biological entity, the action that it performs and Multivariate exploration of a high-dimensional model in the context conditionalities. So, all the processes are defined in systems biology: The delta/notch non-linear dynamic a generic form and are expressed in term of elementary bricks model for cell differentiation of actions found in nature. For the need of a system biology Martens, Harald1; Veflingstad, Siren R.2; Plahte, Erik3; Bertrand, european project BaSysBio for Bacillus subtilis modelling, Dominique4; Martens, Magni5; Omholt, Stig W.6 we have described the central carbon metabolism of Bacillus 1Nofima Food/CIGENE/Dept. of Mathematical Sciences and subtilis using Bioψ to study the glucose/malate shift. From Bioψ Technology, Norwegian University of Life Sciences, Aas, Norway; description which gather in a formal way very rich in biological 2Georgia Institute of Technology, Atlanta, United States; 3CIGENE knowledge, we can generate standard SBML file. From Bioψ (Centre for Integrative Genetics), Norwegian University of Life description we determined “multi-domains” elementary modes of Sciences, Aas, Norway; 4ENITIAA, Nantes, France; 5Nofima the central carbon metabolism. Food (and KVL), Aas, Norway; 6CIGENE (Centre for Integrative Conclusion: Bioψ formalism allows us to build a language for Genetics), Aas, Norway biological functions compatible with biological knowledge which can be applied to elementary flux modes and different kind of A cross-disciplinary explorative approach is presented for system of simulation (cellular automata, multi-agents systems, studying the properties and behaviour of a complex mathematical etc.). model - in this case a high-dimensional nonlinear dynamic model of spatial pattern generation in cell differentiation. The DS2-2-66 approach combines elements from several disciplines in order to provide an overview of the behaviour of the model under a Adaptive modeling mitochondria respiration dynamics wide range of conditions, and in order to discover and quantify via rate equations derived from first principles with unexpected phenomena in the resulting model solutions – with experimentally determinable kinetic parameters limited research work. It includes statistics (reduced factorial Chang, Ivan1; Wallace, Douglas2; Baldi, Pierre3; Letellier, Thierry4 design to probe a high-dimensional parameter space), image 1University of California Irvine, Irvine, United States; 2University of analysis (spatial autocorrelations, clusters etc), sensory science California Irvine, Center for Molecular & Mitochondrial Medicine (verbal concept development, intersubjective sensory descriptive and, Irvine, United States; 3University of California Irvine, School analysis) and chemometrics (Partial Least Squares regression with of Information and Computer Sciences, Irvine, United States; jack-knifing and bilinear graphics). 4Université Victor Segalen-Bordeaux 2, Bordeaux, France, A simplified mathematical model from theoretical biology

Bordeaux, France describes how the differentiation of cells in a 2 D lattice develops Posters due to the mutual interaction between two proteins (Delta and Dedicated Objective: The Oxidative Phosphorylation Pathway of Notch) in any pair of neighbouring cells. This causes a spatial mitochondria is the principle source of energy production in pattern to develop in the lattice. The non-linear differential cells. A model that allows quantitative analysis of the dynamics equations for each cell are simple enough, but the whole system between the rate of electron transfer, work done by the proton of equations for a high number of cells in a 2 D lattice is too translocation, as well as the production of ROS, is essential complex to study, analytically. We performed an explorative for the investigation in the mechanisms of the mitochondria investigation of how the high-dimensional nonlinear dynamic respiration pathologies. We present a new deterministic model model behaves under different conditions - different combinations of mitochondria bioenergetics based on reaction rate laws of model parameter, initial conditions and boundary conditions. derived from non-equilibrium thermodynamics, but containing The results revealed that the high-dimensional non-linear essential kinetic parameters obtainable from experiments in system has a rich repertoire of distinct pattern types, - some of clinical laboratories. Our aim is to provide a platform in which them quite unexpected. Moreover, it revealed how the model a mitochondria system’s behavior dynamics can be uniquely parameters and initial conditions affect these outcomes. In pursuit prescribed through a set of simple experiments, which then can of details, we discovered systematic, but multi-domain bifurcation be used in the study of mitochondrial pathologies. complexities in the state space. Results: Preliminary results have shown that when the kinetic parameters are obtained from experiments for a specific DS2-2-68 mitochondria system, we were able to fit and simulate the behavior of each complex individually, and also able to mimic the Integrative modelling of gene regulatory interactions systems behavior via threshold curves of each complex in the relevant in antirheumatic therapy system. Hecker, Michael1; Goertsches, Robert2; Koczan, Dirk2; Kekow, Conclusions: We have developed a new computational model Joern3; Thiesen, Hans-Juergen2; Guthke, Reinhard1 of mitochondria respiration that is derived from first principles, 1Leibniz Institute for Natural Product Research and Infection validated with experimental data at both component and systems Biology - Hans-Knoell-Institute, Molecular and Applied level, and adaptive to specific mitochondria systems. We hope Microbiology, Jena, Germany; 2University of Rostock, Institute of to further develop the system to eventually aide in the clinical Immunology, Rostock, Germany; 3University of Magdeburg, Clinic diagnostic of mitochondrial pathologies by allowing researchers to of Rheumatology, Vogelsang, Germany perform perturbation studies in-silico. Objective: The biological background of this study is the autoimmune disease rheumatoid arthritis (RA) and the effects of medication by Etanercept (a TNF-alpha blocker). Etanercept therapy in RA patients has proven efficacious, but the underlying molecular mechanisms by which it slows the disease progression remain unclear. A systems-biological approach is suited to unravel the therapeutic effects on gene expression. Results: Applying Affymetrix whole genome microarrays, time- dependent RNA signatures of peripheral blood mononuclear cells from Etanercept treated RA patients were obtained. We analysed the data exploiting information of the GeneAnnot database and identified 86 genes as significantly regulated during first week of therapy. Most of the genes are known to control the body’s immune response. By searching for transcription factor (TF) binding motifs (from Transfac and Jaspar) in the regulatory sequences of these genes we found that their expression is

ICSB 2008 139 regulated by a combination of several TFs. Based on these DS2-2-70 information, we inferred a mathematical model of the modulated gene regulatory network. Thereby, a system of (linear) equations Decoding protein interaction networks was fitted to the data using a L1-regularization based method Reifman, Jaques; Ivanic, Joseph; Wallqvist, Anders (the Lasso algorithm). The estimation of the model parameters is Biotechnology HPC Software Applications Institute, US Army constrained, so that, for instance, each TF may be an activator Medical Research and Materiel Command, Ft. Detrick, United or an inhibitor, but not both. We ended up with a sparse network States model of TF-gene interactions. Regulatory predictions were assessed by literature mining. Objective: Protein-protein interaction (PPI) networks are Conclusions: We developed a method for deriving transcriptional commonly explored for the identification of distinctive biological regulatory networks from gene expression data by integrating traits. Therefore, understanding their underlying network genome sequence information in conjunction with DNA-binding structures is vital to assess the significance of any discovered motifs of TFs. Furthermore, our modelling approach facilitates the features. We recently demonstrated that PPI networks show integration of known protein-protein interactions. When applied degree-weighted (DW) behavior, whereby the probability of on expression profiles of patients treated with Etanercept, the interaction between two proteins is generally proportional network inference uncovers therapeutic effects at the molecular to the product of their numbers of interacting partners, or level and thus provides useful hypotheses about the drugs’ degrees. It was surmised that DW behavior is a characteristic mechanisms of action. of randomness. We expand upon these findings by developing a random, DW, network model, which we contrast against two DS2-2-69 types of PPI networks: i) those determined from a single high- throughput (HT) experiment, and ii) curated compilations, typically Logical modeling of growth factor signaling derived from multiple sources. Samaga, Regina1; Saez-Rodriguez, Julio2; Alexopoulos, Leonidas Results: We find that eight PPI networks determined from single G2; Sorger, Peter K2; Klamt, Steffen1 high-throughput (HT) experiments have global and local properties 1Max-Planck-Institute for Dynamics of Complex Technical that are consistent with this random model. The apparent random Systems, Magdeburg, Germany; 2Harvard Medical School, Dept. connectivity in HT PPI networks is counter-intuitive with respect Dedicated of Systems Biology / MIT, Dept. of Biological Engineering, Boston, to their observed degree distributions, and we resolve this Posters United States discrepancy by introducing a non-network-based model for the evolution of protein degrees, or ‘binding affinities.’ However, in Objective: When studying cell signaling, it is important not to stark contrast, curated networks are found to be inconsistent with confine oneself to single pathways, but also to deal with large- a random connectivity. scale networks that allow a global analysis. Due to lacking Conclusions: The differences between HT-determined and knowledge of kinetic laws and parameters, those large-scale curated PPI networks can be attributed to two main reasons: i) if cellular networks can often only be modeled with qualitative a curated network includes interactions from more than one HT modeling approaches relying solely on the network structure. We data set and the overlap between these sets is small, then the use a Boolean description (logical interaction hypergraphs) for combined network may appear multi-modular, and ii) curated and modeling and analyzing signaling through the EGF-/ErbB receptor high-confidence PPI networks have been manually manipulated family and the HGF receptor c-Met. and, therefore, include biases or preferential influences upon Results: Based on the stoichiometric pathway map of Oda et the protein interactions, which may or may not represent the al. (2005), we built up a large-scale qualitative model (about underlying biology. As many studies infer biological properties 100 species and 200 interactions) of EGFR/ErbB signaling that by contrasting PPI networks against corresponding random comprises the main (canonical) and also the side routes of EGFR/ networks, any discovered inferences must consider the underlying ErbB signaling. Despite its qualitative nature, the model enables construction of such networks. to derive important functional properties and predictions. The model was compared with high-throughput data of primary DS2-2-71 hepatocytes and a transformed cell line (HepG2), where the EGFR/ErbB network was stimulated with different drugs and Simultaneous model discrimination and parameter inhibitors. We demonstrate that our approach enables to uncover estimation in dynamic models of biological systems discrepancies between experimental results and our current Rehberg, Markus1; Rodriguez-Fernandez, Maria2; Egea, Jose A.2; qualitative knowledge and to generate new hypotheses. We also Kremling, Andreas1; R. Banga, Julio2 show how side effects of different drugs can be detected using 1Max-Planck-Institute for Dynamics of Complex Technical this approach. Currently, we are extending the model to the Systems, Systems Biology Group, Magdeburg, Germany; 2IIM- HGF pathway; further pathways such as IGF and TGF-β will be CSIC, Process Engineering Group, Vigo, Spain included in the future. Conclusions: The developed growth factor signaling model Objective: Frequently, a number of alternative models are presented here exemplifies the applicability of our logical modeling proposed to explain a given set of experimental data. Fitting approach for cellular signal transduction networks. It allows separately each of the models from the candidate set and one not only to study the feedback structure or network-wide choosing the best-fitting is the most used procedure. However, dependencies, but also to make predictions on the input-output when the number of possible models is large, the problem behavior of signaling networks and on the qualitative effects of becomes tedious and very costly in terms of computational time. perturbations. Additionally, our approach enables to search for Here we propose a systematic procedure which solves interventions inducing a desired response and is well suited for simultaneously the problem of model discrimination and the analysis and exemplification of huge (discretized) data sets. parameter estimation (model calibration). Our procedure is based Due to the flexible architecture, additional pathways can be easily on a Mixed-Integer Nonlinear Programming (MINLP) formulation integrated in the model. subject to the differential-algebraic (DAEs) model acting as the set of constraints. This problem is non-convex, so a global optimization method based on the scatter search metaheuristic (fSSm) is used to obtain proper solutions. The capabilities of this procedure are illustrated considering the problem of modelling the K+ homeostasis in E. coli. Results: The candidate set consists of 8 different models determined by three different options: regulation of proteolysis, regulation of translation (or lack of them) and different dynamics for one of the feedback loops. These candidates were modelled by means of 3 binary decision variables together with 16 kinetic

140 ICSB 2008 parameters that were treated as continuous variables. systems. It is, however, intriguing that various cellular signaling The MINLP-DAEs problem was successfully solved using fSSm. systems use coupled positive feedback circuits to implement the The quality of the fitting obtained by this methodology was better hysteretic switch. A question then arises about the advantage of than the one reached considering the 8 models separately. using coupled positive feedback circuits instead of simple isolated Moreover, the computational time was reduced in more than one positive feedback for an apparently equivalent hysteretic switch. order of magnitude. Results: Through mathematical simulations, we determined that Conclusions: A simultaneous approach for model discrimination cellular systems with coupled positive feedback circuits show and parameter estimation in nonlinear dynamic models of enhanced hysteretic switching, and thereby can make a more biosystems is proposed here. The approach is based on reliable decision under noisy signaling. a MINLP-DAEs formulation and the use of an efficient and Conclusions: As most intracellular processes are accompanied robust global optimization method. A case study considering with intrinsic noises, important cellular decisions such as homeostasis in E. coli was successfully solved by this approach, differentiation and apoptosis are required to be highly robust to such showing a significant reduction of computation time with respect noises. The coupled positive feedback circuits might have been to the standard approach. evolutionarily acquired to make a correct cell fate decision through enhanced hysteretic switching under noisy cellular environments. DS2-2-72 DS2-2-74 Exploring the parameter properties of the Delta-Notch model using GEMANOVA on sensory data from generated A biochemical programming language for gene regulation images networks Isaeva, Julia1; Saebo, Solve1; Wyller, John A.1; Liland, Kristian H.1; Salama, Rafik Faergestad, Ellen Mosleth2; Bro, Rasmus3; Martens, Harald4 University of Birmingham, Birmingham, United Kingdom 1Norwegian University of Life Sciences (UMB), Ås, Norway; 2Nofima food, Matforsk AS, Ås, Norway;3 University of Introduction: A gene regulation modelling grammar and Copenhagen, Copenhagen, Denmark; 4Centre for Integrative language is proposed, that abstracts the rules governing the Genetics, Ås, Norway transcription and translation of transcription units that can be complicated enough if to be regulated by many factors. The Objective:The Delta-Notch model is a well established grammars proposed so far are, a) focused more on capturing the mathematical model for describing cell-pattern development [1]. rules rather than representing it in a more usable or easy to learn It models the pattern-generating ability of two signalling proteins and reusable manner , b) not rendering any biochemically valid

(Delta and Notch) controlling cell differentiation in either a 1D cell output of their representation which is a major importance of their Posters chain or a 2D lattice. We consider here the 5-parameter non- representation for simulation. Dedicated linear dynamic model for 2D hexagonal lattices describing how Objective: Our research is mainly focused on targeting those each cell interacts with its six neighbours. two aspects by representing the grammars in a programming The pattern generating properties of this model were studied language approach which is: (i) compiled to chemical equations ( by Veflingstad et al.[2]. They chose two levels for each of the in a suitable biochemical language e.g. SBML so that it interfaces five parameters: low and high. In addition two levels for each of with SBML-ready simulation engines) using fewer simple three initial conditions were chosen. A 27-2 fractional design was statements which saves the user from representing the regulation run, and upon convergence of the models this gave 32 different of a transcription unit with many hundreds of reactions, (ii) pictures which were shown to 11 judges comprising a sensory captures rules in a machine understandable language that can be panel. They used 12 pre-defined attributes to describe the easily verified, maintained, simple to use (either for a modeller or pictures. biologist) and reuse (i.e. reusing functions and structures of gene To explore influence of each level of those parameters on the regulation in the form of a programming library). images of patterns (and hence, on the scores from the judges), Conclusion: The language and the libraries proposed in this Veflingstad et al.[2], assumed a linear model, and PLS-ANOVA work has been used to efficiently model complex gene regulation was used for estimation. The ANOVA model revealed that four networks in fairly small programs which represent hundreds of of the model parameters were the main contributors to pattern chemical reactions. The networks modelled include melR-melAB formation. The authors also found significant low-order interaction transcription units of Escherichia coli and the korAB transcription effects between these. unit in IncP-1 plasmids. Finally we represent the ways the Results:In order to explore the interaction effects further we language can be extended to include more complex biochemical analyzed the same data using a GEMANOVA model [3,4], but behaviours and constructs that allow for user-extensibility of its considering only the four most important factors, one attribute function. at a time and the average score across all judges, i.e. a four- dimensional array (2x2x2x2). A forward selection scheme coupled DS2-2-75 with bootstrapping was used to find the optimal GEMANOVA models. This analysis yielded interesting interaction effects for Role identification in complex biological processes applied several of the attributes. The results are here compared with the to the cell cycle model of fission yeast PLS-ANOVA results previously published. Zeemering, Stef; Peeters, Ralf; Westra, Ronald Conclusions:The GEMANOVA model is well suited for finding Maastricht University, Department of Mathematics, MICC, higher order interactions, which often are problematic to find by Maastricht, Netherlands standard ANOVA analysis. Objective: Automated role identification in complex biological DS2-2-73 processes based on linearized interaction matrices using notions and techniques from systems and control theory. We aim to The role of coupled positive feedback loops and hysteresis distinguish between activating and inactivating roles of regulatory in biological networks process entities. These roles may vary during different process Kim, Jeong-Rae; Cho, Kwang-Hyun phases along with changes in the internal feedback control Korea Advanced Institute of Science and Technology (KAIST), mechanisms. The developed method is applied to the well-known Department of Bio and Brain Engineering, Daejeon, Republic of Tyson-Novak model of the cell division cycle of fission yeast. Korea Results: The discrete mass switch in the original cell cycle model has been successfully replaced by a continuous switch that Objective: Hysteresis is found in many physical systems, and produces similar cyclic behavior. This transforms the model into a a hysteretic switch has been used for various mechanical and completely continuous model that is accessible to a larger array electrical systems. Such a hysteretic switch can be created using of mathematical tools. The roles and role transitions of the model a single positive feedback circuit, as often used in engineering entities have been computed by analyzing the structure of a

ICSB 2008 141 linearized version of the model at several points on the limit cycle. organic acids acetate and butyrate. During the transition phase C. These roles and role transitions coincide to a large extent with the acetobutylicum switches towards the generation of the solvents

known cell cycle phases (G1, S, G2 and M) and checkpoints (G1/S butanol and acetone as dominant fermentation products, a

and (G2/M). process called solventogenic shift. The results of the experiments Conclusions: The developed procedure for automated role show that the crucial prerequisite for the induction of this identification is able to detect cell cycle phases and checkpoints metabolic change is the external pH. The definite inducement of in the Tyson-Novak model of the cell division cycle of fission yeast. the switch is not figured out completely yet. Two approaches have been tested, based on the values or the Objective: Shinto et al. published in 2007 a first kinetic model signs of the linearized model interactions. Both approaches yield to describe the dynamic behaviour of the metabolic pathway promising results, most notably the detection of the strong central of C. acetobutylicum. This model neglects the known pH-

role of Cdc13T. An important characteristic of the procedure is dependent behaviour of C. acetobutylicum. Based on this model that it does not simply register peaks or switches, but that it looks we developed a sophisticated model by focusing on the pH- at the role of an entity in the overall process instead. We believe dependency and including additional biological information such that the procedure can serve as a useful tool in the identification as inhibition and activation processes as well as a pH-dependent of phases and checkpoints in complex biological processes in switching factor. The aim of this work was describing the which the underlying dynamics are largely unknown. metabolic switch as a function of pH-value. Results: We constructed a new model based on the Shinto DS2-2-76 approach and attached additional control- and regulation mechanisms which induce a change in the dynamics. After Microarray Data clustering using Monte Carlo simulation of cutting out the energetic and Clostridia specific improbable Potts spin model reactions, attaching pH-dependent terms as well as regarding Torres, Luis1; Rosu, Haret2; González, José3; González, Adonaí2 stoichiometry, we obtained a more realistic view into the biphasic 1Universidad Iberoamericana Leon, Ciencias Básicas, León, Gto., metabolism of C. acetobutylicum. Mexico; 2IPICyT, Materiales Avanzados, San luis Potosí, Mexico; Conclusion: The metabolic switch of C. acetobutylicum from 3Centro Universitario Lagos, UG, Modelación y Matemáticas acidogenisis to solventogenisis depends on the internal pH Aplicadas, Jalisco, Mexico which can be controlled by the external pH in experiments. The Dedicated inclusion of a pH-dependent kinetics is therefore necessary to Posters Objective: describe the behaviour of the bacteria for different pH-values. DNA microarrays allow us to explore a major subset or all genes The consideration of related activation and inhibition processes is of an organism. The metabolic architecture for less complex important for the modeling of C. acetobutylicum. organisms, such as Escherichia coli, the biochemical network has been described in much detail. Here we investigate the clustering DS2-2-78 of such gene networks by applying the Potts spin model to the gene expression data point by introducing an interaction between Global sensitivity analysis methods for dynamic models in neighboring points, whose strength is a decreasing function of the systems biology distance between neighbours. Rodriguez-Fernandez, Maria; R. Banga, Julio Results: IIM-CSIC, Process Engineering Group, Vigo, Spain Potts model has been studied extensively for many years. The basic spin variable s can take one of q integer values: s = 1, 2… Objective: Developing suitable dynamic models of biochemical q. In a magnetic model, the Potts spins are located at points pathways is a key issue in Systems Biology. Parameter

vi that reside on or off the sites of some lattice. Pairs of spins identification is therefore a critical aspect where the analysis of the associated with points i and j are coupled by an interaction of model identifiability plays an important role. The study of model

strength Jij>0. Denoting by S a configuration of the system, S identifiability aims to determine whether the unknown parameters

= {si}i=1,N , the energy of such a configuration is given by the can be uniquely estimated from the available experiments. As the

Hamiltonian, H (s) = ∑ Ji,j (1-δsi,sj), si= 1,……q , where number of parameters and the complexity of the models increase, the notation stands for neighboring sites vi and vj. The current methods for testing structural identifiability become contribution of a pair to H is 0 when si= sj, that is, when the inapplicable, so methods mostly based on local sensitivities

two spins are aligned, and is Jij >0 otherwise. The Hamiltonian is are frequently used. Although these methods are valid for linear very similar to other energy functions used in neural systems. cases, or when the value of the parameters is known, they can be Conclusions: misleading for the general nonlinear case. In this work, each spin variable represents a q-state gene Global sensitivity analysis is presented here as a robust alternative expression level. Monte Carlo simulation methods (Swendsen- for this type of analysis, detecting non influential parameters Wang and Metropolis) overcome this problem by generating and their interactions. The performance of this methodology a characteristic subset of configurations, which are used as a is illustrated with a benchmark dynamic model describing statistical sample. They are based on the notion of importance a biochemical pathway which has been previously used in sampling, in which a set of spin configurations (S1, S2…Sk) is parameter estimation studies. generated according to the Boltzmann probability distribution. So Results: The derivative based global sensitivity measures (DGSM) far, we have defined the Potts model, the various thermodynamic presented by Kucherenko, Rodriguez-Fernandez, Pantelides & function that one measures in this case, and the numerical Shah (to appear in Reliab, Eng. Syst. Safety) for explicit models method used to measure these quantities. We also make the were here extended in order to be able to handle differential- simulations of these concepts on the gene clustering data. algebraic equations (DAEs). A parameter ranking based on these measures was established and compared with the one based on DS2-2-77 local sensitivities. Moreover, a global correlation matrix built from the DGSM pH-Dependent Modeling of ABE Fermentation in was computed providing more complete information about Clostridium acetobutylicum interactions than the local one obtained from the Fisher Haus, Sylvia; Millat, Thomas; Wolkenhauer, Olaf Information Matrix. University of Rostock, Systems Biology and Bioinformatics, Conclusions: DGSM for a model of a biochemical pathway with Rostock, Germany 36 kinetic parameters and 8 states were contrasted with the well known Sobol’ indices method showing similar behaviour but with Clostridium acetobutylicum is a commercial valuable bacterium, a considerable reduction of the computational time. first isolated from corn in 1912 by Chaim Weizmann. The The differences found between the local and global ranking metabolism of this obligate anaerobe, non-pathogenic bacterium demonstrate that this one is more accurate for prioritisation of is characterized by the Acetone-Butanol-Ethanol (ABE) parameters. Moreover, the global correlation matrix allowed us fermentation. Exponentially growing cells mainly produce the to fix the pairs of parameters highly correlated, simplifying the

142 ICSB 2008 parameter estimation problem. the same one with different constraints) could provide different models that represent the same data, which is an expected DS2-2-79 feature of such methods. Experiments would then have to be designed to discriminate between competing models, in a way The validity of the reaction-diffusion master equation in 1, 2 that ‘closes the loop’ between modelling and experiment. and 3 dimensions Sjöberg, Paul; Elf, Johan DS2-2-81 Uppsala University, Dept. of Cell & Molecular Biology, Uppsala, Sweden Stochastic model of translation Thiel, Marco1; Stansfield, Ian2; Grebogi, Celso1; Romano, M Objective: The reaction-diffusion master equation (RDME) Carmen3 describes a reaction-diffusion process mesoscopically by 1University of Aberdeen, Physics, Aberdeen, United Kingdom; dividing the reaction volume into a number of subvolumes, each 2University of Aberdeen, Institute of Medical Sciences, Aberdeen, homogeneous in the distribution of molecules. We study how the United Kingdom; 3University of Aberdeen, Physics and Institute of reaction rates in the master equation formalism depends on the Medical Sciences, Aberdeen, United Kingdom microscopic reaction rates of the Smoluchowski framework and the size of the subvolumes. Objective: We propose a model for the process of translation, Results: If the length scale of the subvolume is large compared i.e., how ribosomes translate the messenger RNA molecules to the reaction radius, the rates of reactions in the subvolume into proteins that can be utilised by the cell for a huge variety of are well defined in three dimensions and independent of different processes. the subvolume size. We show how the discretization that Results: In order to model translation, we propose a simple is introduced by the division into subvolumes affects the stochastic model based on the totally asymmetric exclusion mesoscopic reaction rates in lower dimensions. process. We focus on the role that different distributions of It is well known that the mesocopic association rate in a nucleotides, i.e., different mRNA sequences, play in the maximal homogeneous three dimensional volume depends on the diffusion flow or production rate of proteins that can be achieved. We then constant, the microscopic reaction rate and the reaction radius. relate the rich dynamical behaviour generated by the model to the Since a dissociation event in practice is a series of dissociations different biological functions of the computed proteins. and reassociations, the mesoscopic dissociation rate will depend Conclusions: The specific position of codons in the mRNA on the diffusion rate as well. sequence plays a very important role for the translation rate Conclusion: The mesoscopic reaction rates of RDME in one of the corresponding protein. It does influence not only the and two dimensions do not only depend on the diffusion rate, maximal translation rate that can be achieved, but also how the Posters the microscopic reaction rates and the reaction radius, but also translation rate changes in dependence on different environmental Dedicated the division into subvolumes. By deriving mesoscopic rates conditions. from the microscopic rates RDME can be used in a way that is consistent with the microscopic level of description in one and DS2-2-82 two dimensions. Mathematical model of the Lux luminescence system in the DS2-2-80 terrestrial bacterium Photorhabdus luminescens Welham, Patricia Determining interconnections in biochemical networks University of Birmingham, Systems Biology, Birmingham, United using Linear Programming Kingdom August, Elias Oxford university, Oxford, United Kingdom Objective: To construct a mathematical model of the Lux luminescence system, governed by the operon luxCDABE in the Objective: The aim of this work is to determine the connectivity terrestrial bacterium Photorhabdus luminescens, using a set of of biochemical networks from time-series experimental data. The coupled ordinary differential equations, and to test the system on objective is to increase our understanding of the functionality of time series and stationary data from published papers. complex biochemical systems. Results: The model is in good agreement with the published Results: We present a methodology for efficient and robust data. Metabolic control analysis demonstrates that control of the determination of the interaction topology of biochemical systems system lies mainly with the aldehyde recycling pathway (LuxE using time-series data collected from experiments (with often only and LuxC) and that the aldehyde substrate of the light emitting a few data points available), under the assumption that these reaction, catalysed by LuxAB, has the most effect on the steady networks are sparse; i.e., they have much less edges than the state velocity of the system. The rate at which light is produced full graph with the same vertex set. To achieve this, we minimise in the steady state model shows a low sensitivity to changes in the 1-norm of the decision variables while constraints keep kinetic parameter values to those measured in other species of experimental data in close Euler fit to the network dynamics thus luminescent bacteria, demonstrating the robustness of the Lux putting more emphasis on determining the interconnection than system. on the closeness of fit. We have reformulated the problem as a Conclusions: This model will have value in the interpretation of Linear Programme for which efficient algorithms are available even Lux data when used as a reporter in time-course gene expression for the case of uncertainties in data or model parameters. experiments. First, we consider a system in which the interconnection strengths (unknown parameters) enter the system dynamics in an affine DS2-2-83 way. For example, chemical reaction networks with mass action kinetics are systems that have such a structure. We illustrate Disconnectivity approach as a new dimension in the the ability of our method to identify a chemical reaction network topological analysis of regulatory networks first through a numerical example, followed by applying it to Potapov, Anatolij; Goemann, Björn; Wingender, Edgar real experimental data of the glycolytic pathway of L. lactis: Medical School, Georg August University of Göttingen, the network structure thus obtained compares well with what Department of Bioinformatics, Göttingen, Germany is believed to be the known structure. Second, we extend our approach to the case of gene regulatory networks, in which Objective: There is a gap between purely theoretical studies the system dynamics are much more complicated. A numerical of the topology of large regulatory networks and the practical example showing the effectiveness of our method is given. traditions and interests of experimentalists. While a theoretician Conclusions: The presented methodology was able to predict emphasizes the global characterization of regulatory systems, the connectivity structure of two reaction networks and a gene an experimentalist focuses on the role of distinct molecules and regulatory network but for some links. Different methods (or genes in regulation. To bridge the gap between these opposite

ICSB 2008 143 approaches, one needs to combine ‘general’ with ‘particular’ their association with various biological contexts is central for properties and translate abstract topological features of large obtaining systems-level understanding of biological processes. systems into testable characteristics of individual components. We investigate co-expression patterns of the genes in a protein Results: We propose a new topological parameter - the pairwise interaction network across a versatile collection of normal human disconnectivity index of a network element - that is capable of tissues. The objectives are two-fold. The first goal is to identify such bridging. The index quantifies how crucial an individual and characterize co-regulated gene groups that have alternative element is for sustaining the communication ability between co-expression patterns across biological conditions. The second connected pairs of vertices in a network that is displayed as task is to describe relationships between the co-expression a directed graph. Such an element might be a vertex (e.g., patterns and biological context. Differently from previous co- molecules, genes), an edge (e.g., reactions, interactions), as well expression studies that investigate predefined gene groups as a group of vertices and/or edges. The index evaluates the or focus on pairwise relationships between genes, we extend topological redundancy of regulatory paths connecting different co-expression analysis into larger gene groups using network parts of a network and sensitivity (robustness) of this network to information, and describe the differences in co-expression the presence of each individual element. The method has been between various biological conditions. applied to the analysis of the transcription regulatory networks Results: We apply a recently developed subspace clustering from E. coli and S. cerevisiae, the neuronal connectivity network technique that utilizes network information to guide the grouping of C. elegans, and the mammalian TLR4 signal transduction of the genes to identify functionally coherent local regions in network. The program for evaluating the pairwise disconnectivity the network that have coordinated expression across various indices is available at http://www.bioinf.med.uni-goettingen.de/ biological conditions. Such functionally coherent subnetworks services/. are often associated with known biological processes with higher Conclusions: The biological importance of the suggested specificity than in alternative methods. The distinct co-expression approach relies on its capacity to quantitatively evaluate the patterns of the subnetworks are in many cases associated with topological significance of every vertex, edge, their groups and specific biological contexts such as immune system, muscles, or combinations in the context of all other elements in a given the brain. regulatory network: that is how a given network can be regulated Conclusions: Our findings highlight the context-sensitivity of by means of its reorganization, i.e., removing an element and gene regulatory mechanisms and the varying role of biological Dedicated restoring the element. processes in changing cell-biological environments. We Posters demonstrate how the activation of the biological processes can DS2-2-84 change in varying contexts, and also suggest novel functional hypotheses for gene groups having distinct co-activation patterns Effects of multiple parameter variations on biological that cannot be explained by known functional annotations. system behaviors Wang, Ruiqi; Chen, Luonan DS2-2-86 Institute of Systems Biology, Shanghai, China Giant glial cell: New insight through mechanism-based Objective: Biological processes often involve many parameters, modeling and the importance of these parameters in determining system Brazhe, Alexey1; Postnov, Dmitry2; Brazhe, Nadezda3; behaviors must be assessed in order to gain deep insight Sosnovtseva, Olga3; Ryazanova, Ludmila2; Mosekilde, Erik3 into designing principles of living organisms. However, many 1Denmark Technical University, Institute for Physics, Lyngby, parameters driving the behavior of biochemical circuits vary Denmark; 2Saratov State University, Physical faculty, Saratov, extensively and are therefore difficult to analyze the effects of Russian Federation; 3Denmark Technical University, Institute for individual parameters in a systematic way. Physics, BioSim group, Lyngby, Denmark Results: By decomposing a closed-loop system which may be monotone or non-monotone into some open-loop but We propose a generalized mathematical model of a small neural- monotone subsystens or modules, we can use the input- glial ensemble. The model incorporates the subunits of the output characteristics to study the effects of multiple parameter tripartite synapse, which includes presynaptic and postsynaptic variations on system behaviors in a systematic way. The neurons, and a glial cell. Mechanism-based model is validated for proposed approach can be used to not only positive feedback a tripartite synapse consisting of P- and R-neurons and a giant loop systems but also systems with negative feedback loops. In glial cell in the ganglia of medical leech (Hirudo medicinalis), which those systems with negative feedback loops, a discreet map can is a useful object for experimental neurophysiology studies in situ. be used to build the correspondence between the closed and We describe the two main pathways of the glial cell activation: open loop systems. (i) via IP3 production and Ca2+ release from the endoplasmic Conclusion: The main ideas are illustrated through a positive reticulum and (ii) via increase of the extracellular potassium feedback loop system, i.e. a five-variable system Mos/MAPK concentration, glia depolarization, and the activation of the kinase p42 MAPK cascade and a system with negative voltage-dependent Ca2+ channels. We suggest that the second feedback loops, i.e. the Goldbeter’s oscillatory model although pathway is more significant for establishing the positive feedback the proposed technique is valid for a general class of biological in the glutamate release which in its turn is critical for the self- systems. The results show that the proposed approach is very sustained activity of the postsynaptic neuron. This mechanism effective to analyze effects of multiple parameter variations on differs from the mechanisms in the astrocyte-neuron signaling system-wide behaviors in a systematic way. The approach can reported by previous authors. also help us design informative experiments and gain deep insight into regulation mechanisms of living organisms. DS2-2-87

DS2-2-85 Uncovering transcriptional and post-transcriptional regulations mechanisms using sequential logic model Identification of alternative co-expression patterns in a Piras, Vincent1; Sasidharan, Kalesh1; Zhen Xuan, Yeo2; Giuliani, protein interaction network Alessandro3; Tomita, Masaru1; Selvarajoo, Kumar1; Tsuchiya, Lahti, Leo1; Knuuttila, Juha E.A.2; Kaski, Sami1 Masa1 1TKK Helsinki University of Technology, Department of Information 1Keio University, Institute for Advanced Biosciences, Tsuruoka, and Computer Science, Espoo, Finland; 2University of Helsinki, Japan; 2Genome Institute of Singapore, Singapore, Singapore; Neuroscience Centre, Helsinki, Finland 3Istituto Superiore di Sanita’, Rome, Italy

Objective: Cell-biological processes are regulated in a complex Objective: The study of Gene Regulatory Networks (GRNs) network of interactions between genes and their products. is a challenging area since we now realize global and holistic Understanding the co-activation patterns of the genes and understanding of the dynamic regulation of gene expression

144 ICSB 2008 is necessary. Studies initially focused on the transcriptional DS2-2-90 regulation by Transcription Factors (TFs). Recently post- transcriptional regulation mediated by micro-RNAs (miRNAs) Design and analysis of a reduced ODE model of IKK has been discovered and shown to play an important role in phosphorylation following stimulation with IL-1 and UVB gene regulation. Moreover, messenger RNA (mRNA) decay Witt, Johannes1; Schwarzwälder, Eva2; Barisic, Sandra2; Kulms, machineries add another regulatory layer. All these processes Dagmar2; Sawodny, Oliver1; Sauter, Thomas1 work in parallel and dynamically cooperate with each other to 1University of Stuttgart, Institute for System Dynamics, Stuttgart, maintain the desired mRNAs or proteins levels in the cell. Such Germany; 2University of Stuttgart, Institute for Cell Biology and complexity makes the interpretation of experimental data difficult Immunology, Stuttgart, Germany and it is therefore necessary to develop comprehensive dynamical computational approaches, integrating both transcriptional and NF-κB dependent enhancement of UVB-induced apoptosis has post-transcriptional regulations to support the understanding of recently been reported upon co-stimulation of human epithelial the mechanistic properties of these regulations. cells with IL-1. This phenomenon coincides with a sustained NF- Results: We previously developed Sequential Logic Model κB activity caused by sustained IKK phosphorylation. (SLM) to decipher the dynamic transcriptional control of gene Systems biology can help to gain further insight into the signaling expression. Here, we further extend SLM to model mRNA decay processes leading to IKK phosphorylation. One major focus in and miRNA-directed mRNA cleavage. Translational regulation by modeling should thereby be on model simplicity; models should miRNAs is also taken in account, as in GRNs context, mRNAs not be more complex than required for data explanation. can be translated into TFs to regulate transcription of other genes We design a strictly reduced ordinary differential equation model (regulatory loop). We also describe delay models since they without feedback for IL-1 induced phosphorylation of IKK, using affect the dynamics of GRNs. Finally, using different synthetic the PottersWheel toolbox. Essential processes concerning the GRNs involving TFs, miRNAs, and feedback loops, we perform IL-1 receptor, IKK and the IKK phosphatase PP2A are included simulation of our models to measure their dynamical behavior in the model. We show that, in contrast to existing models, IKK upon stimulation (e.g. activation of TFs), and perform in-silico can be considered decoupled from the downstream proteins. knock-outs (e.g. miRNAs or TFs knock out) to determine the Simulation results agree well with the experimental data for high critical regulatory molecules and thresholds such as minimum IL-1 stimulation with and without UVB costimulation. We use the miRNA concentrations. model to show that constitutive PP2A activity is consistent with Conclusion: SLM provides a fruitful non-parametric and model- the experimental data and a possible hypothesis for IKK inhibition. independent integrated approach to reveal control mechanisms In addition, we suggest that PP2A is indirectly rather than directly and analyze the complex interactions and cooperation between influenced by UVB radiation.

Transcription factors, miRNAs and mRNA decay machineries to For low IL-1 doses, delayed IKK phosphorylation is observed, Posters regulate gene expression. which requires a model extension consistent with the previous Dedicated model. We hypothesize that TRAF6 autoubiquitination is DS2-2-89 responsible for a positive feedback upstream of IKK. The thus extended model agrees well with all experimental data sets, Escherichia coli respiratory adaptation agent-based model including data for PP2A knockdown cells. Identifiability analysis Maleki-Dizaji, Saeedeh1; Holcombe, Mike1; Green, Jeff2; Rolfe, reveals that the model allows precise predictions concerning Matt3; Poole, Robert3; Ghraham, Alison2 PP2A deactivation and IKK dephosphorylation rate constants. 1The University of Sheffield, Computer Science, Sheffield, United This study clearly demonstrates that the advantages of reduced Kingdom; 2The University of Sheffield, Department of Molecular models are enhanced process understanding, parameter Biology and Biotechnology, Sheffield, United Kingdom;3 The identifiability and hypotheses testing. University of Sheffield, Department of Molecular Biology and Biotechnology, Sheffield, United Kingdom DS2-2-91

Objective: Escherichia coli is an organism able to successfully A divide-and-conquer framework for global parameter grow at a wide range of oxygen levels, with many components optimization of biochemical models of the oxygen utilisation and sensing machinery already known; Kotte, Oliver; Heinemann, Matthias however, no comprehensive knowledge exists of their regulation ETH Zurich, Institute of Molecular Systems Biology, Zurich, and how they interact with each other. Instead, current knowledge Switzerland comes from experiments carried out using different bacterial strains, growth conditions and measurement techniques that Objective: Recent years have seen an increase in both the examine isolated parts of the system. The work detailed here number and complexity of kinetic models of biochemical systems. uses these existing “fragmentary” datasets to develop an E. coli These models generally involve many uncertain parameters, respiratory adaptation Agent-based model using FLAME (Flexible which are usually estimated using a global optimization method. Large-scale Agent-based Modelling Environment). Typically, such optimization problems are large-dimensional, Results: The simple agent-based model of E. coli oxygen underdetermined and non-convex, posing substantial challenges regulation is based upon a biological model consisting of to the employed optimization algorithm e.g. in terms of molecular oxygen (O2), cytochrome bo and bd-I terminal oxidases computational time, and to the modeler in terms of transparency. (mRNA and protein), FNR and ArcA. In the agent-based model, We therefore developed a framework that can lead to a drastic each of these biological components is represented by an agent, decrease in computational time, and to a drastic increase in which can communicate with other agents via message passing. problem transparency. The function of each agent is defined by a set of rules (conditions) Results: The here presented framework divides the global constructed according to the biological model, which allows optimization problem into multiple independent problems implementation of the activity (behaviour) of each agent and is of smaller dimension size, which are solved individually and defined by the context of the agent. are subsequently joined to the global solution. We present Conclusions: The model predicts that many components of mathematical proof that this approach guarantees a global the system should behave as is already known, especially FNR optimum, given certain conditions on the measurement data, activity, ArcA activity and cytochrome oxidase mRNA levels. the distribution of parameters, and the model structure. We Estimates of cytochrome oxidase protein levels did not correlate argue that real-world biochemical models are likely to fulfill as expected, indicating that a lack of current knowledge may be these conditions. We illustrate the applicability of the method limiting the behaviour of the model. This work is the first attempt by estimating the parameters of a small model system, which to model oxygen regulation in E. coli using an agent-based includes both metabolic and transcriptional regulation layers, with model, and suggests that useful predictions can extrapolated respect to published experimental data. from previously published evidence produced under different Conclusions: The divide-and-conquer framework can provide conditions. a feasible alternative to competing global optimization methods,

ICSB 2008 145 with important advantages. First, the achieved reduction in included. The final step is to convert the specified information to required computational time can enable the estimation of a mathematical model, where the states describe intuitive, but very large-dimensional parameter sets. Second, the increase possibly overlapping, pools of possible protein configurations. in problem transparency can lead to further insights into the Importantly, our reduced models are not approximations of modeled system. Third, discrepancies between the measurement corresponding state-space models, even though we show data and the assumed model structure are immediately identified examples where the number of states in public models of real and localized, thereby linking the often separated problems of signalling systems have been reduced by orders of magnitude. parameter and model structure uncertainty. Our approach therefore is a major step forward towards solving the combinatorial explosion problem. DS2-2-92 DS2-2-94 In silico analysis of robust developmental modules Munteanu, Andreea; Solé, Ricard V. An integrated analysis of the metabolic and regulatory PRBB - University Pompeu Fabra, Barcelona, Spain networks of Escherichia coli K12 Carneiro, Sónia1; Lourenço, Anália1; Mendes, Rui2; Rocha, Objective: Advances in evolutionary developmental biology have Miguel2; Ferreira, Eugénio1; Rocha, Isabel1 revealed that small numbers of genes form robust functional 1IBB - Institute for Biotechnology and Bioengineering,Center modules, hierarchically reused throughout development. In the of Biological Engineering, Braga, Portugal; 2Departament of present work, we analyzed by means of numerical simulations of Informatics / CCTC - University of Minho, Braga, Portugal diverse network topologies the evolution of such small modules towards higher diversity and robustness. Consequentially, the Objective: The major goal of this work is the integrated analysis object of study consists in the mapping between genotype (gene of metabolic and genetic networks of E. coli K12, by performing network topology) and the resultant phenotype (observable gene network topology analysis, motif finding and the simulation of cell expression profile). The latter is the factor upon which natural behaviour. We use a novel logic framework, which provides users selection acts and thus the driver of continuous evolution. The set with a powerful language to query information using both first and of all genotypes, their resultant phenotypes and their associated second order predicates. Dedicated fitness defines a fitness landscape. The nature and evolutionary Results: The overall network integrates data from publicly Posters implications of the fitness landscape still remain key topics in available repositories, namely EcoCyc, BRENDA and RegulonDB, evolutionary developmental biology. and Palsson’s manually-curated stoichiometric model1. The Results: Our computer simulations employ Boolean networks, metabolic level comprises all known chemical reactions catalyzed with the Boolean nature defined by the ON/OFF states of the by enzymes, with the corresponding reactants and products, constituent genes. The small size of the studied networks allowed which may also act as metabolic regulators (inhibitors and an exhaustive analysis of the entire fitness landscape and the activators). Genes, promoters, transcription factors, transcriptional extent of its neutrality. This analysis as well as simulations of inducers and sigma factors, constitute the elements of the gene evolutionary processes uncover a set of genetic interactions regulation level. Results from topological analysis confirmed producing robust and diverse phenotypes. We single out the previous studies on E. coli models, highlighting the importance of distinctive features of these networks responsible for their stability several biomolecules in the cellular metabolism. As expected, σ against environmental and structural perturbations. All these 70, CRP transcriptional factor and metabolic cofactors, such as robust genotypes can be related to the key mechanism of lateral ATP, NADH and others, are major network hubs. The integration inhibition for which a cell of a given type inhibits its neighbors to of both metabolic and genetic regulation in the same network, keep them from adopting the same type. allowed to perform a novel integrated analysis and to reach Conclusions: The current study provides clues on possible interesting results, namely the identification of the pathways in underlying principles of network assemblage and evolution which one or the other type of regulation is dominant. Additionally, in developmental processes. The uncovered class of robust it was possible to observe that in several pathways those modules shows that these networks are structurally robust: regulation mechanisms are used together to provide a range of many mutations are neutral and provide back-up mechanisms or metabolic responses, e.g. in citric acid cycle and glycolysis. alternative pathways. In addition, lateral inhibition appears to be Furthermore, results revealed the intrinsic complexity of reaction not only a generic form of creating ON-OFF spatial patterns, but a networks, involving several structural motifs. The uncovering mechanism of producing robustness. motifs illustrated a tight control of metabolism. Conclusions: An integrated view of the metabolic and regulatory DS2-2-93 networks is of paramount importance in the elucidation of the mechanisms of regulation in different cellular processes. A contingency-based approach to mapping and modelling Moreover, in silico strain improvement procedures through of cellular networks metabolic engineering design may benefit by considering Cedersund, Gunnar1; Funahashi, Akria2; Kitano, Hiroaki3; Krantz, simultaneously this different types of cellular regulations. Marcus4 1. Feist, A. M. et al. Mol. Syst. Biol. 3, 121 (2007). 1Linköping University, Linköping, Sweden; 2Keio University, Yokohama, Japan; 3The Systems Biology Institute, Tokyo, Japan; DS2-2-95 4University of Gothenburg, Gothenburg, Sweden Towards a new mathematical model of the cell cycle of The combinatorial explosion problem for signalling systems is Saccharomyces cerevisiae that the number of states in a model increases exponentially Barberis, Matteo1; Vanoni, Marco2; Klipp, Edda1; Alberghina, Lilia2 with the number of independent interactions. This rapid increase 1Max Planck Institute for Molecular Genetics, OWL - Otto is such a severe limitation that it often has been pointed as the Warburg Laboratories, Berlin, Germany; 2University of Milano- major obstacle for modelling of real-size signalling systems. Here, Bicocca, Department of Biotechnology and Biosciences, Milan, we present a new strategy for documenting existing knowledge Italy about signalling systems, and for generating corresponding computational models, in a way that eliminates the biggest part Objective: The cell cycle is a complex biological process whose of the combinatorial explosion problem. The approach contains function requires the interaction of a large number of components three major steps. First, the fundamentally different types of (about 15% of the gene products of budding yeast), whose reactions are described in a graphical representation. Second, understanding can be strongly improved by a systems biology the contingencies among these interactions are specified in a approach. Budding yeast is an established model organism for contingency matrix. This representation of existing knowledge cell cycle studies, and several mathematical models have been has the important advantage compared to classical state- proposed for its cell cycle [Chen et al, Mol Biol Cell, 2004; Chen space graphs that no unsupported assumptions need to be et al, Mol Biol Cell, 2000]. An area of major relevance in cell cycle

146 ICSB 2008 control is the G1/S transition, characterized by the requirement eventually, in transformed cells appears relevant for future of a critical cell size, PS, to enter S phase [Jorgensen and Tyers, perspectives in clinical application. Curr Biol, 2004]. We presented a mathematical model including a more complete set of players of the G1/S transition to give a DS2-2-97 rational explanation to the setting of PS [Barberis M et al, PLoS Comput Biol, 2007]. Sic1 can regulate the fundamental events in the budding Here we integrate the G1/S model within a previously described yeast cell cycle cell cycle model [Alberghina et al, Oncogene, 2001] that takes Barberis, Matteo; Klipp, Edda into account essential features of the cell cycle regulation, namely Max Planck Institute for Molecular Genetics, OWL - Otto Warburg the phase-specific accumulation of cylins and formation of Cdk- Laboratories, Berlin, Germany cyclin complexes. Results: The model considers the main features of the G1/S Objective: The modeling of the G1/S transition in budding transition [Barberis M et al, PLoS Comput Biol, 2007], notably yeast has shown that a systems biology analysis can highlight the explicit compartmentalization that is essential to assign the implications of the cell size in the events driving DNA the specific function to all the proteins and protein complexes replication1. The balance between Cdk1-Clb5,6 kinase — that relevant for the yeast dynamics, together with the specific role induces firing of the DNA replication origins2,3 — and Sic1 cyclin- of the phase-specific Cdk-cyclin complexes in different stages dependent inhibitor is crucial to regulate DNA replication events. of the cell cycle. Positive and negative feedback loops, acting at Here we analyze the involvement of Sic1 in the control of the different level in the control of the Cdk-cyclin complexes, have replication events. Budding yeast cell cycle is driven by periodical been included. The model is implemented by a set of ordinary fluctuations in the activity of the Cdk1 kinase, and different differential equations (ODEs) that describe the temporal change of pools of Clb cyclins associate with Cdk1 in a temporal cascade the concentration of the involved proteins and protein complexes to coordinate the cell cycle’s critical events. This originates the and includes equations describing the temporal change of the behavior widely known as “waves of cyclins”. The transition from concentrations of the involved components, and changes in the low to high Clb activity is triggered by the degradation of Sic1. compartment volumes. Here we address the intriguing question whether Sic1 plays a Conclusions: The model will be used in connection with regulative role in this process, and how this coordinated fashion is experimental approaches (see abstract by Coccetti and achieved. colleagues) as a framework to analyze connection of different Results: We provide a rational explanation of the activation of signaling pathwaysis with the cell cycle machinery. DNA replication initiation due to the Sic1 localization, and its role as promoter of Cdk1-Clb5 to start DNA replication4, as

DS2-2-96 experimentally reported5. Comparing mathematical models that Posters differ in the regulatory interactions in which Sic1 is involved, Dedicated A systems biology analysis of the entry of resting we propose a specific sequence of events that reproduces the mammalian cells into S phase successive activation of the phase-specific Cdk1-Clb complexes. Barberis, Matteo1; Alfieri, Roberta2; Chiaradonna, Ferdinando3; This suggests that Sic1 can offer a feedback loop regulation to Gaglio, Daniela3; Milanesi, Luciano2; Vanoni, Marco3; Klipp, Edda1; trigger the “waves of cyclins”, and the timing of their appearance. Alberghina, Lilia3 Dynamic and stochastic simulations support this hypothesis. 1Max Planck Institute for Molecular Genetics, OWL - Otto Conclusions: The understanding of biological processes has Warburg Laboratories, Berlin, Germany; 2CNR - Institute for strongly improved by the construction of mathematical models Biomedical Technology, Milan, Italy; 3University of Milano-Bicocca, able to predict the dynamics of the process in a variety of Department of Biotechnology and Biosciences, Milan, Italy conditions. Here we propose that the order of cell cycle’s critical events is determined by the specificity of regulatory interactions. Objective: In order to maintain mass homeostasis, cell 1. Barberis M et al (2007) PLoS Comput. Biology 3, 649-666. 2. proliferation requires tight coordination of mass accumulation, Bell SP and Dutta A (2002) Annu. Rev. Biochem. 71, 333-374. a continuous process, and DNA replication and cell division, 3. Takeda DY and Dutta A (2005) Oncogene 24, 2827-2843. 4. which occur only once during each cell cycle. In tumor cells the Barberis M and Klipp E (2007) Genome Inform. 18, 85-99. 5. balance between intracellular and extracellular signals and the Rossi RL et al (2005) Cell Cycle 4, 1798-1807. control of the cell cycle is deranged [Hanahan and Weinberg, Cell, 2000]. Systems biology aims to understand cell cycle DS2-2-98 regulation in organisms of different evolutionary complexity, and various groups have reported different cell cycle models [Chen Stability and feedback regulation in yeast central et al, Mol Biol Cell, 2004; Swat et al, Bioinformatics 2004; Qu et metabolism al, Am J Physiol Cell Physiol, 2003; Haberichter et al, Mol Syst Steuer, Ralf1; Martins, Ana M.2 Biol, 2007; Csikasz-Nagy et al, Biophys J, 2006]. Published 1Manchester Interdisciplinary Biocentre, Manchester, United G1/S networks of mammalian cells do not consider the nucleo/ Kingdom; 2Virginia Bioinformatics Institute - Virginia Tech, cytoplasmic localization relevant to simulate the dynamics of S Blacksburg, United States phase entry, while in neoplastic cells altered localization of key cell cycle players has been reported. Here we present a mathematical Objectives: One of the most challenging goals of systems model of the G1/S transition in normal mammalian fibroblasts, biology is to develop detailed kinetic models of cellular that accounts for differential protein localization, by a combined metabolism. However, the construction of large-scale dynamic experimental/computational approach. models is as yet hampered by incomplete knowledge about Results: Mouse fibroblasts were analyzed during a synchronous the kinetic properties of the involved enzymes and membrane re-entry in cell cycle after serum starvation/re-feeding, and transporters. Motivated by increasing experimental accessibility of the level of the major players determined. Design of the G1/S cellular characteristics, such as metabolic flux and concentrations network took into account literature data and, as largely accepted of intermediates, we describe a general framework for analyzing [Csikasz-Nagy et al, Biophys J, 2006], evolutionary conservation large-scale models of metabolic processes - a framework which of the basic structure of the network. The mathematical approach is not based on any explicit knowledge of enzyme-kinetic rate was as described for budding yeast [Barberis et al, PLoS equations and the associated parameter values. Comput Biol, 2007]. The model computed time courses for total, Results: Each stationary metabolic state, characterized by given cytoplasmic and nuclear concentration of each species. The metabolite concentrations and a flux distribution, is associated nucleo/cytoplasmic ratio of the proteins- a major determinant with a unique spectrum of dynamic properties, as defined by of the G1/S transition according to our model and not used for the ensemble of all possible kinetic models consistent with parameter estimation - was used for model validation. the respective state. Our framework builds upon a statistical Conclusions: The G1/S transition is altered in many human exploration of the ensemble of possible models. With the cancers. The identification of the G1/S network in normal and, method already successfully applied on medium-scale systems,

ICSB 2008 147 we present a dynamic description of the S. cerevisiae central in guinea pig ventricular cells. We then compared a stability of BP metabolism under different environmental conditions. Our primary cells with that of sinoatrial (SA) node pacemaker against possible interest is the organization and optimality of allosteric regulation to fluctuation in plasma electrolyte concentration. In general, BP maintain the function and stability of the network. cells was less stable than SA node pacemaker cells against Conclusions: Evaluating the dynamic capabilities of a metabolic fluctuation in external Na+ and K+ concentration. The difference in system, our conclusions are: (i) Despite the prevailing picture behavior between SA node cell and BP cell was most remarkable of flux distributions as ‘steady-states’, metabolic networks under hyperkalemia. We consider that the less stability of BP cell

generically have rich dynamics and a stable solution represents was probable caused by insufficient depolarization viaI Na. an infinitesimal fraction of parameter space only. (ii) Metabolic Conclusions: Induced BP activity was successfully simulated states and feasible flux distributions can show drastic differences with E-CELL SE. Evaluation of the stability of BP showed that BP in their stability properties, with profound consequences for cells are generally less stable than SA node cells against possible biotechnological applications. (iii) The role of each reaction and, fluctuation in plasma electrolyte concentration. in particular, the role of allosteric feedback regulation in the stabilization of the metabolic network can be quantified. It is DS2-2-103 anticipated that biotechnological modifications of metabolism bring about concomitant changes in dynamic properties, making Evolving partially unknown genetic regulators networks our framework a major step towards determining the feasibility of Wegner, Katja1; Knabe, Johannes F.2; Schilstra, Maria J.1; intended modifications. Nehaniv, Chrytopher L.2 1University of Hertfordshire, Biocomputation Research Group, DS2-2-100 Hatfield, United Kingdom;2 University of Hertfordshire, Adaptive Systems Research Group, Hatfield, United Kingdom Complex systems modelling of stem cell fate decisions Julianne, Halley1; Winkler, David2; Burden, Frank2 Objective:Genetic regulatory networks (GRNs) form the 1CSIRO, Molecular and Health Technologies, Victoria, Australia; basic layer of control for many processes in living cells. The 2CSIRO, Molecular and Health Technologies, Clayton South MDC, structure and dynamics of GRNs are often complex and not fully Australia understood, evolutionary algorithms (EAs) may be used to attempt Dedicated to fill the gaps. EAs are appropriate to find ‘good’ (not necessarily Posters Despite intense experimental research, the ways in which the ‘best’) solutions to problems with a large search space. stem cells integrate and process information remains unclear, We are developing an extension to our tool NetBuilder’ (which compromising our ability to manipulate their behaviour. Although can be used to create and simulate layered Petri net models of animal body plans are highly reproducible (sometimes to the GRNs) that employs an EA to evolve GRNs that exhibit specific level of single cells in simpler animals), it is doubtful that the full responses to given input signals. complexity of the adult mammalian body plan is reducible to Results: The new algorithm developed for NetBuilder’ is based instructions specified in DNA, without other pattern formation on our experiences with an EA that uses a binary representation mechanisms at play. In particular, interaction between selection of a genome with a fixed number of genes. In contrast to existing processes and mechanisms of self-organization account for an approaches, our EA does not need to start with a randomized awesome array of pattern in biology, and it is becoming clear that population but can take existing knowledge into account. cellular therapies could benefit substantially from appreciation Therefore, the first individual can be a Petri net that contains all of this interaction. We explore the role of self-organization in regulatory elements that are known to be part of the network. haematopoietic and embryonic stem cell decision making. Our The EA uses randomized selection, crossover, duplication and work comprises a conceptual framework to integrate theories of mutation to change populations and improve the fitness over self-organization with what is known of the regulatory architecture a number of generations. The fitness is the agreement of the of stem cells, and a computational model to transform this simulated and the target output behaviour, with a penalty for understanding into a rigorous methodology capable of interacting the size of the network to make sure that smaller networks are with biochemical experiments. The conceptual model has a favoured by the selection process. number of components. Stochastic gene expression plays an Conclusions: The EA is able to adjust the network structure important role in priming gene regulatory network modules close and parameters to find a GRN with a certain target function to criticality where they are more sensitive to external signals. (e.g. sine or step). In the near future, we will add the functionality The interplay between stochastic priming and these signals to use experimental data as input or target behaviour to make determines how the signalling cascade determines stem cell fate. hypothesis about missing network connections or parameters in The experimental evidence for these model components, and the partially unknown GRNs. results of preliminary calculations of network behaviour, will be discussed. DS2-2-104

DS2-2-101 A method for analyzing robustness and application to two models of the cyanobacterial circadian cycle Evaluation of artificial biological pacemaker induced via Hafner, Marc1; Koeppl, Heinz1; Wagner, Andreas2 1 suppression of Kir2.1 channels: A simulation study Ecole Polythechnique Fédérale de Lausanne, Laboratory of Takeuchi, Maria; Sano, Hitomi; Kumamoto, Hiromi; Naito, Nonlinear Systems, Lausanne, Switzerland; 2University of Zurich, Yasuhiro; Tomita, Masaru Department of Biochemistry, Zurich, Switzerland Keio University, Institute for Advanced Biosciences, Fujisawa, Japan Objective: We propose a new method to analyze and quantify the robustness of biochemical models. Current analyses are Objective: Recent advances have demonstrated several based either on local or global properties. Our method tries to approaches to creating artificial biological pacemaker (BP) cells merge these two approaches by looking at sets of models and from quiescent ventricular cells via adenoviral gene transfer, as finding correlations between regions of the parameter volume an alternative to electoronic pacemakers that have potential risk and local properties. Two models of the cyanobacterial circadian of pacemaker pocket infections. The present study evaluated cycle (Mehra et al. 2006 and Rust et al. 2007) are chosen as case a stability of BP cells against various changes in external study. environment, such as plasma electrolyte concentration. Results: The first step of the method is to define relevant Results: We simulated suppression of Kir2.1 channel current properties for the output. For each property we define a

(IK1) in a comprehensive electrophysiological cardiomyocyte consistent range in which we expect the output to be. For model, known as the Kyoto model, on the basis of the E-CELL the cyanobacterial circadian cycle the output is the total Simulation Environment (SE). Our simulation reproduced well the concentration of phosphorylated KaiC. The period, the peak value

reported behavior of the induced BP activity via suppression of IK1 and the amplitude are the relevant properties and the consistent

148 ICSB 2008 ranges are defined around the nominal values. DS2-2-106 With a Metropolis algorithm, we find sets of parameters that give results in the consistent ranges. We apply a principal component Robust model simplifications in multiscale systems analysis on these consistent sets to get the eigendirections and Radulescu, Ovidiu1; Gorban, Alexander2; Zinovyev, Andrei3; the distributions of the volume of the consistent parameter sets. Lilienbaum, Alain4 The next step is to refine the analysis by looking at different 1University of Rennes, CNRS UMR 6025, Rennes, France; local robustness properties of the consistent parameter sets. 2University of Leicester, Center for Mathematical Modeling, The first one is cycle stability for which we calculate the Floquet Leicester, United Kingdom; 3Institut Curie, Bioinformatics Lab, multipliers. Next we analyze robustness to fluctuation of the Paris, France; 4Universit e Paris 7, CNRS URA 2115, Paris, total concentration of Kai by applying random perturbations and France checking if the model still produces consistent results. Finally we measure robustness to molecular noise by calculating the signal- Objective: There are many reasons for simplifying systems noise ratio of stochastic simulations. biology models. Models contain unnecessary complexity Conclusion: The results show that Mehra’s model has a strong which conceals design principles, and renders analysis difficult. constraint in the parameter volume. This is less the case for Identification of critical targets, parameters, regulations become Rust’s model and its normalized parameter volume is larger. easier for reduced models. Many pathways work together, Rust’s model is then more robust as parameters can fluctuate therefore we need to know how to couple models in order to in a larger volume. With the local analyses, we find that some obtain larger ones. To be compared, or to be integrated into parameters rule the robustness properties. Therefore some sets larger ones, models must be simplified to a common level of should be discarded if we take in account local robustness. complexity. Thus, model comparison and model coupling needs Especially the robustness to molecular noise in the model from model reduction. Understanding of physiology could rely on Mehra is controlled by two parameters. By comparison, the having a hierarchy of robust simplifications. model from Rust is more robust to molecular noise. Results: We propose model reduction algorithms for reactions networks with hierarchies of time scales, that can be DS2-2-105 systematically applied in mathematical modeling of biological pathways. The algorithms are based on 1) generalized theory of Integration of genetic and enzymatic controls in metabolic the limiting step that we developed earlier; 2) searching dominant pathways solutions of quasi-stationarity equations; 3) averaging, i.e. Bekkal Brikci, Fadia1; Goelzer, Anne2; Fromion, Vincent2 introducing approximate conserved quantities. In all cases, we 1Institut de Recherche en Agronomie(INRA), Mathématiques, obtain robust model simplifications. The simplified model contains 2 Informatique et Génome, Jouy en Josas, France; INRA, Jouy en much smaller number of components and parameters, but has Posters Josas, France approximately the same dynamical behavior as a complex model. Dedicated The reduction algorithm also identifies the critical parameters. Objectives: The bacterial metabolic machinery and its regulation We systematically apply model reduction to a detailed NFkB are a complex system and involve many components: metabolites pathway model developed by us earlier, and obtain a hierarchy and enzymes. The reconstruction of the regulatory networks of of models of decreasing complexity having the same dynamical Bacillus subtilis including the chemical reactions, the known gene properties. We demonstrate that more complex models in this transcriptional, translational and post-translational regulations and hierarchy are more robust with respect to variation of parameters controls of enzymatic activities (see [1]) allowed us to identify the that provides an insight into biological network design principles. main regulation structures of metabolic pathways. Then, in order We show that these model properties are related to general to break down the intrinsic complexity of such a huge system and mathematical results in measure concentration theory, and to reveal somewhat an order, we propose a new suitable notion suggest a classification of robust behaviors. of subsystems (modules) using mathematical dynamic models Conclusion: Robust simplifications of multiscale models are associated to these structures. new and powerful tools for the study of systems biology models. Results: The definition of a suitable subsystem structure and the This research is supported by “Mathematical theory of biological analysis of their associated equilibrium allowed us to understand robustness with experimental applications in cancer systems the key role of some elements of the metabolic pathways: biology” project. Visit http://bioinfo.curie.fr/projects/matbrac to • The key role of irreversible steps, obtain the references. • The key role of the regulated enzymes and the metabolite associated to their regulations, DS2-2-107 • The connection between modules with global regulators. And thus to understand some interesting biological questions: CaliBayes: Bayesian calibration of biological simulation • Why in a linear metabolic pathway both the genetic and models enzymatic activity controls are often located at the first enzyme Boys, Richard; Henderson, Daniel; Kirkwood, Tom; Wilkinson, level? What is the role of isoenzymes? Darren; Wolski, Eryk; Wu, Jake • What is the impact of so-called “pleiotropic” regulators on Newcastle University, Newcastle upon Tyne, United Kingdom metabolic pathways? • Can the enzymes, metabolites concentrations and fluxes be Objective: The development of a flexible software infrastructure qualitatively predicted? for statistical calibration, analysis and validation of systems We present here some answer elements and their illustrations biology computer models using time course experimental data. through the study of two examples: the purine synthesis in Results: Using Java, web services (WS), a service oriented Bacillus subtilis and the complex interconnection of the threonine architecture (SOA), the systems biology markup language and isoleucine syntheses in Escherichia coli. (SBML), a standard format for time course data and a standard Conclusion: These dynamical systems combine both enzymatic biological simulator API, it is possible to build a generic software and genetic levels through the integration of the metabolic infrastructure for Bayesian parameter estimation and associated network feedback on the genetic network by metabolites. The analysis that is largely independent of the nature of the experiment study of their associated equilibrium allows us to deduce some that generates the data, the biological model, and the simulation biological properties, and finally the ability to break down the huge engine. The computationally intensive inference algorithms rely metabolic network on suitable set of subsystems. on Markov chain Monte Carlo (MCMC) technology and can be References: [1]. A. Goelzer et al.: Reconstruction and analysis applied to both deterministic and stochastic models. Further, for of the genetic and metabolic regulatory networks of the central large and/or complex models, the procedure can be split into metabolism of Bacillus subtilis, BMC 2008. two stages. The first stage consists of building a fast statistical emulator of the model. Although very computationally intensive, this stage is easy to distribute across a computer cluster. The second stage then consists of running the inference algorithm

ICSB 2008 149 using the emulator in place of the true simulation model. This regulation using expression data remains difficult, mainly for stage is also computationally intensive and is difficult to parallelise two reasons: (i) there is still a lack of microRNA expression effectively, but the use of a fast emulator renders it practical. The data with adequate time resolution, and (ii) the inhibitory effect technology will be illustrated using a real model and associated of microRNAs on translation stays invisible on the mRNA experimental data. level. Here we study gene regulation under the influence of Conclusions: A SOA for the Bayesian analysis of SBML models microRNAs controlling differentiation of ES cells into endoderm using time course data is shown to be an effective and scalable and mesoderm lineages. We uncover microRNA-driven gene solution to model calibration and validation. The software is free regulation by measuring the indirect influence of an intronic and portable, and can be used stand-alone via web-services or a microRNA on a microRNA-TF-target-gene motif. simple web interface. Results: Starting from two known TFs regulating endoderm development, Foxa2 and Sox17 [Matsuura, Stem Cell, 2006], DS2-2-108 we identify several potential intronic microRNAs targeting the corresponding transcripts using a weighted combination of A first analysis of the central carbon pathway of Bacillus the available microRNA-target-site prediction tools. To quantify subtilis: Integration of both genetic and enzymatic levels in microRNA expression we take the host-gene expression of the regulations embedded intronic microRNAs as proxy [Tsang, Molecular Cell, Goelzer, Anne1; Fromion, Vincent2 2007]. From the expression data of the TF targets, we identify a 1Institut de Recherche en Agronomie, Mathématiques, qualitative regulatory motif that stabilizes the differentiation into Informatique et Génome, Jouy en Josas, France; 2INRA, France, mesoderm lineage consisting of the mesoderm specific host gene France Mest, the embedded microRNA mmu-miR-335 and the TFs. Based on this motif we develop a quantitative model to study the Objective: A challenging problem in systems biology consists dynamical behavior of the system. in understanding the regulation and the global coordination of Conclusion: Using microarray expression data of differentiating metabolic pathways of bacteria in response to environmental ES cells, we identify a regulatory system including an intronic changes, at the cell scale. Recently we identified the main microRNA that stabilizes differentiation during development. regulation structures of metabolic pathways through the manually Through dynamical modeling we identify a bistable system that Dedicated curated reconstruction of the regulatory networks of Bacillus can switch between undifferentiated mesoderm and endoderm Posters subtilis [1]. We noticed a strong interplay between the metabolic state starting from an undifferentiated cell. This sheds light into pathways and the genetic regulations: 13% of 456 metabolites the potential cellular mechanisms driving ES cell differentiation. are involved in genetic regulations and 53% of the genes involved in metabolic pathways are directly regulated by a genetic regulator DS2-2-110 under the control of a metabolite. Surprisingly, few studies centred on metabolic pathways have considered this coupling Capturing metabolite flow in biological systems through between the genetic and metabolic regulatory networks. kinetic radiotracer simulations Results: Using a mathematical dynamic model, we illustrate the Subbanna, Nagesh1; Padiadpu, Jyothi1; Zeevi, Yehoshua2; necessity of integrating both genetic and metabolic regulations, Chandra, Nagasuma1 through the study of glycolysis, during exponential growth on 1Indian Institute of Science, Bioinformatics, Bangalore, India; glucose. Through a sensitivity analysis of this model, we identified 2Technion - Israel Institute of Technology, Electrical Engineering, (i) the modules in the central carbon metabolism using the Haifa, Israel definition in [1]; (ii) key reactions in glycolysis (iii) four metabolite pools as key signals, and the nature of information they integrate; Modelling metabolic pathways using labelled reactants is (iv) the respective role of genetic and enzymatic regulations. We invaluable to understanding flow of metabolites within and also predicted the behaviour of the enzymes and metabolites between pathways, particularly in understanding differences that concentrations, of the fluxes. Thanks to this analysis, the initial arise due to a disease such as cancer. Here, we carry out kinetic model have been simplified and integrated into a simulator in simulations of NAD metabolic pathways containing labelled Matlab, which allows us to prove the interest of the proposed starting isotopes, and track their flow by a new probabilistic analysis for increasing our understanding of regulations. algorithm that determines the most probable routes in the Conclusions: We obtained a semi-quantitative behaviour in network. Starting with known reactions and their rates described simulation in agreement with published data, both at the genetic by Michaelis-Menten first order differential equations, we and the metabolic levels. However, the gap between semi represent the reactants using nodes of a graph, and the reactions quantitative and quantitative predictions still requires a large set of using edges weighted by the reaction rates. experiments to identify all parameters included in the model. We compute the probabilities of state transitions (reactions) References: using normalised reaction rates and obtain a Markov probability [1]. A. Goelzer et al.: Reconstruction and analysis of the genetic matrix. The change of probabilities over time can be computed and metabolic regulatory networks of the central metabolism of easily using a combination of recursive procedures, Chapman- Bacillus subtilis. In press 2008. Kolmogorov equations and matrix multiplications. The most probable route, which denotes the most important enzymes DS2-2-109 and their concentrations for the generation of NAD at any given instant of time can be directly deduced from the probability matrix Identifying and modeling a microRNA influenced regulatory using backtracking, while concentrations of reactants in the NAD system in stem cell development pathway are obtained by multiplying the initial concentrations with Lutter, Dominik1; Marr, Carsten2; Uetzmann, Lena3; Lickert, Heiko3; the probability matrix. Our computational efficiency is high due to Theis, Fabian J.2 the recursive computation of probabilities. 1Helmholtz Zentrum Muenchen, IBIS, München, Germany; With the aid of simulations, we establish that formate and NADP 2Helmholtz Zentrum Muenchen, Institute for Bioinformatics and are formed three and 2.5 times faster in cancer cells, compared Systems Biology, München, Germany; 3Helmholtz Zentrum to normal cells. The reaction paths involving NAD, NADP, and Muenchen, Stem Cell Institute, München, Germany NMN, among others, are examined and compared in normal and cancer cells. The differences between the healthy and normal Objective: The differentiation of embryonic stem (ES) cells cells are highlighted. Finally, the correctness of our results is during gastrulation into the cell types of the three germ layers is tested by comparing our results with experimental data. Our controlled by multiple interacting gene regulatory mechanisms. technique can be extended to all systems whose concentrations In addition to transcription factor (TF) driven regulations, there is over time are known. strong evidence that microRNAs play an important role during stem cell self maintenance and differentiation [Hornstein, Nat Genet, 2006]. The identification of microRNA-controlled gene

150 ICSB 2008 DS2-2-111 exhibits a reduced fitness and a shorter replicative life span. The model confirms the findings in budding yeast and moreover Modelling the mTOR signalling pathway in dendritic protein simulations suggest that asymmetrical distribution of damage synthesis increases the fitness of the cell population as a whole at both low Jain, Pragati; Bhalla, Upinder S. and high damage propagation rates and pushes the upper limits National Centre for Biological Sciences (NCBS), Bangalore, India for how much damage the system can endure before entering clonal senescence. Thus, we suggest that “sibling-specific” aging Objective: Long-term potentiation (LTP) and long-term in unicellular systems may have evolved as a byproduct of the depression (LTD) are mechanisms important for memory. The strong selection for damage segregation during cytokinesis, and mTOR pathway modulates protein synthesis crucial for LTP and may be more common than previously anticipated. LTD. Three major inputs (BDNF, mGluR and Ca) play a role in controlling protein production. Our aim is to design an in-silico DS2-2-113 pathway which will represent dendritic protein synthesis. We analyse the pathway for protein production rate and its sensitivity Noise propagation and sensitivity in stochastic reaction to activation of various molecules. Further, we use this model to networks find out the contribution of various inputs to protein production Kim, Kyung H.1; Qian, Hong2; Sauro, Herbert M.1 and how production rate varies in response to LTP- and LTD- 1University of Washington, Department of Bioengineering, Seattle, inducing activity patterns. United States; 2University of Washington, Department of Applied Results: We have developed a detailed mass-action model of Mathematics, Seattle, United States BDNF and activity dependent regulation of protein synthesis, with specific application to synaptic plasticity. It involves binding Objective: We have developed a theoretical approach for of BDNF to TrKB receptors and triggering multiple downstream investigating the relationship between noise propagation and molecules. Our model includes six newly developed sub-models system sensitivities in biochemical networks. In particular: (A) We or modules, for different pathways that regulate protein synthesis. investigate how noise propagations affect system sensitivities We validated our model by testing each module individually depending on different network topologies and regulation and the merged model against published experimental data. patterns. (B) We also investigate how sensitivities are related The model replicates data that show that BDNF significantly to one another by proposing a stochastic version of metabolic phosphorylates and activates key players necessary for protein control analysis (MCA). translation. It predicts that both LTP and LTD stimuli elicit an Results: (A) We have investigated how noise propagations affect increase in protein synthesis, which is greatly enhanced when mean rates of reaction. The mean rates have been shown to

BDNF is simultaneously present. We find that the rate of protein depend on how concentration fluctuations cause the fluctuations Posters production is very sensitive to S6K phosphorylation which is of the propensity functions [Gomez, Verghese. J. Chem. Phys. Dedicated regulated by several kinases, and this regulation may lead to 126: 024109 (2007)]. Based on this result, we have shown expression of different proteins. Despite several putative positive that the sensitivities can be enhanced in one region of system feedback loops in the model, it seems unlikely to be bistable. parameter values by reducing sensitivities in another region. Conclusion: Our simulations show that presence of both We have applied this sensitivity compensation effect, e.g., to BDNF and MAPK is essential for increase in synthesis of enhance the amplification of a concentration detector designed proteins. Protein synthesis is selectively regulated by the level by using incoherent feed-forward reaction network. (B) We have of intermediate active molecules. The behaviour of the model is also shown that MCA can be extended for stochastic reaction consistent with experimental data. systems by providing new stochastic versions of summation and connectivity theorems. DS2-2-112 Conclusions: (A) We propose the sensitivity compensation effect as a mechanism that can provide qualitative understanding and Selective benefits of damage partitioning in unicellular prediction of how noise propagations affect system functions. (B) systems; effects on robustness, fitness and aging We have also demonstrated that the noise propagations prevent Cvijovic, Marija1; Erjavec, Nika2; Nyström, Thomas2; Klipp, Edda3 stochastic version of the connectivity theorems from connecting 1Max Planck Institute for Molecular Genetics, Computational between local system sensitivities and global sensitivities. We Systems Biology Group, Berlin, Germany; 2Göteborg University, have discussed that our MCA-like theorems can be improved Göteborg, Sweden; 3Max Planck Institute for Molecular Genetics, further by adopting modular analysis of the noise propagation Berlin, Germany [Tanase-Nicola, et al. PRL 97:068102 (2006)].

Objective: Cytokinesis in unicellular organisms sometimes entails DS2-2-114 a division of labour between cells leading to lineage-specific aging. To investigate the potential benefits of asymmetrical Improving network inference models for gene regulatory cytokines, we created a mathematical model to simulate the networks robustness and fitness of dividing systems displaying different Burroughs, Nigel1; Juarez, Miguel2; Morrissey, Edward2 degrees of damage segregation and size asymmetries. 1University of Warwick, Systems Biology centre, Coventry, United Results: The model suggests that systems dividing Kingdom; 2Warwick University, Systems Biology centre, Coventry, asymmetrically (size-wise) or displaying damage segregation are United Kingdom more robust than fully symmetrical systems, i.e. can withstand higher degrees of damage before entering clonal senescence. Objective: Analysis of gene expression data requires a number Both size and damage asymmetries resulted in a separation of of simplifying assumptions in order to limit model complexity. It is the population into a rejuvenating and an aging lineage. When unclear which assumptions are justifiable, and how they affect the considering population fitness, a system producing different-sized inferred networks. We address this issue in a combined simulation progeny, like budding yeast, is predicted to benefit from damage and experimental data study. retention only at high damage propagation rates. In contrast, Results: We investigate Bayesian inference of dynamic Bayesian the fitness of a system of equal-sized progeny is enhanced by networks from microarray data, exploring a number of model damage segregation regardless of damage propagation rates improvements including methods for imposing sparsity in the suggesting that damage partitioning may provide an evolutionary network to time augmentation. We developed Markov chain advantage also in systems dividing by binary fission. Using Monte Carlo algorithms for model inference with typical run times S. pombe as a model, we demonstrate experimentally that varying from 0.5-5mins per gene on 100 gene networks and damaged, oxidized, proteins are unevenly partitioned during quantify model performance by a series of measures, including cytokinesis and that the damage-enriched sibling suffers from a ROC curves (simulated data) and Bayes factors. prolonged generation time and an accelerated aging. Conclusions: We demonstrate that use of Gaussian Bayesian Conclusion: We demonstrate that the damage-enriched cell networks as an inference model can be improved by careful

ICSB 2008 151 consideration of the key features of the underlying model that it is disease genes solely from PPI network topology. Our method approximating. generalizes the degree of a node, measuring the number of edges that a node “touches”, into a vector of 73 “graphlet degrees” DS2-2-115 (GDs), measuring the number of graphlets that a node “touches”, where graphlets are small connected non-isomorphic induced Describing saturation phenomena in models of regulated subgraphs of a large network. We call this vector the “GD- metabolic networks signature” of the node. We design a GD-signature-based node Tenazinha, Nuno; Vinga, Susana similarity measure that quantifies the similarity of nodes’ local Instituto de Engenharia de Sistemas e Computadores - network neighborhoods. Investigação e Desenvolvimento, Lisbon, Portugal Conclusion: We apply our new GD-signature-based network clustering method to PPI networks of yeast and human to Objective: The dynamic modeling of biochemical networks uncover biological function of unclassified proteins and validate constitutes a major challenge in systems biology. Saturation our predictions in the literature. Furthermore, we show a striking phenomena are ubiquitous features of these systems given the similarity of GD-signatures of cancer genes and demonstrate inherent biochemical processes such as enzymatic cooperativity how to use our technique for cancer/disease gene identification. and gene regulation events. Nonetheless until recently there Our method can thus provide valuable guidelines for future was no straightforward and systematic way of addressing these experimental research. aspects when using approximate modeling methods such as power-laws or the lin-log and the log(linear) formalisms. DS2-2-117 We herein compare three different approaches of including saturations in models of regulated metabolism: power-laws Shrinkage of extracellular space during enhanced neuronal in their piecewise formulation [1], mixed modeling with Hill activity. A modeling approach functions embedded in power-laws, and the recently proposed Østby, Ivar1; Øyehaug, Leiv1; Einevoll, Gaute T.2; Omholt, Stig W.1 Saturable and Cooperative formalism [2]. A benchmark artificial 1Norwegian University of Life Sciences (UMB), Centre for metabolic network [3] that includes feedback inhibition and Integrative Genetics (CIGENE), Ås, Norway; 2Norwegian University feedforward activation regulation is used to establish comparisons of Life Sciences (UMB), Department of Mathematical Sciences Dedicated regarding the mathematical maneuverability of the formalisms, and Technology, Ås, Norway Posters their associated computational and numerical aspects and the possible biological insights given by the different approaches. Objective Neuronal activity triggers release of neurotransmitters Results: The piecewice power-law method has the advantages and ions from the neuron, which in turn changes the composition widely acknowledged for these approaches although some of ion species in the extracellular space (ECS) between neurons interpretability is lost for the additional model parameters. The and surrounding astrocytes. This is followed by ~30% shrinkage mixed modeling with embedded Hill functions is useful when of the ECS and swelling of the adjacent astrocyte. We present a priori understanding exists on the saturable variables of the a dynamic model based on information on important astrocyte system and can have the advantages of the power-law formalism membrane processes and which accounts for experimental through recasting methodologies. The Saturable and Cooperative data on the shrinkage phenomenon and the clearance of formalism allows better insights into the dynamic features of the ECS potassium. Imposing empirically based constraints on networks through numerical simulation results. various model configurations, the study aims at evaluating the Conclusions: The further development of mathematical significance of transport processes that may be essential in formalisms and numerical tools that appropriately accommodate activity-induced ECS shrinkage. saturation phenomena is of major importance especially when Results Neuronal release of potassium and uptake of sodium creating integrative models for metabolic and regulatory networks. during stimulation, astrocyte uptake in passive channels, action References: [1] Savageau M., Math Biosci 2002. 180: 237-253 of the sodium-potassium pump and action of water channels [2] Sorribas A., Hernández-Bermejo B., Vilaprinyo E., Alves R., (aquaporins) together seem sufficient for generating ECS Biotechnol Bioeng 2007. 97(5): 1259-1277 shrinkage as such. The action of Na+/K+/Cl- and Na+/HCO3— [3] Marino S., Voit EO., J Bioinform Comp Biol 2006. 4(3): 665- -cotransporters appears critical for achieving observed quantitative 691 levels of ECS shrinkage, and for explaining ECS and astrocyte ion concentrations observed during neuronal stimulation. The DS2-2-116 presence of these cotransporters in combination implies enhanced ability to reproduce observed ion concentrations. Knocking out Uncovering disease genes and function via graphlet degree the Na+/K+/Cl- cotransporter and reducing the permeability of signatures aquaporin water channels, the model predicts only subtle changes Przulj, Natasa; Milenkovic, Tijana in ECS shrinkage compared to wild type. University of California, Irvine, Computer Science, Irvine, United Conclusions The present study (i) indicates the importance States of the Na+/K+/Cl-- and Na+/HCO3— -cotransporters, and, in particular these two in combination, in terms of reproducing Objective: Proteins are essential macromolecules of life and empirically observed ion concentrations and levels of ECS understanding their function and their role in disease is important. shrinkage and (ii) predicts that relevant gene knockouts have However, the number of functionally unclassified proteins is limited effect on ECS shrinkage. The study identifies the large even for simple, well studied organisms. Moreover, it is need for measuring ECS shrinkage, membrane potential, ion still unclear in what cellular states serious diseases, such as concentrations and identifying water flow direction in wild type cancer, occur. In protein-protein interaction (PPI) networks, and gene knockout individuals. nodes correspond to proteins and edges represent interactions between them. Since proteins aggregate to perform a function, DS2-2-118 PPI networks by definition reflect the systems-level representation of interconnected nature of biological processes. Therefore, A collective intelligence approach to modelling intelligent analyzing structural properties of PPI networks may provide useful cellular organisation insights into function of individual proteins, protein complexes, Periyasamy, Sathish1; Kille, Peter2; Gray, Alex3 larger cellular machines, and complex diseases. Finding the 1Cardiff University, Cardiff School of Computer Science, relationship between PPI network topology and biological function Cardiff, United Kingdom; 2Cardiff University, Cardiff School of and disease remains one of the most challenging problems in the Biosciences, Cardiff, United Kingdom; 3Cardiff University, Cardiff post-genomic systems-biology era. School of Computer Science, Cardiff, United Kingdom Results: We present a systems-level proteome-scale method that addresses this challenge. Our method is capable of not only Objective: The aim of this approach is to adopt principles of uncovering protein function, but also detecting cancer and other Collective Intelligence (CI) in representing the intelligent cellular

152 ICSB 2008 organisation. A biological cell consists of various molecular development in data modeling techniques. High-speed multi- species confined to distinct locations in the cell. These molecules wavelength bio-spectroscopy methods such as NIR and FTIR, have no centralised control and use a distributed problem solving high-dimensional MS and NMR metabolomics spectrometry and strategy to sustain the cellular organisation. The main features other relatively low-cost profiling technologies are advancing of CI that we strive to implement are limited communication rapidly. Combined with robotics, they may increasingly provide and interactions, large number of interacting entities with limited systems biologists with sufficient chemical detail at sufficient contact and some globally efficient, emergent or self-organising temporal resolution to enable biological control structures to be behaviour. We implement these features using an Agent Based discovered, characterized and verified from empirical data. The Modelling and Simulation (ABMS) environment to capture the present work concerns a new data modeling approach intended cellular level phenomena. Further we intend to progress towards to match the measuring revolution in biology. The objective is to developing an in silico based synthetic minimal biological cell. contribute to a better balance between prior causal assumptions Results: A prototype model using the above approach has vs. information present in multivariate time series data. Thereby, been implemented using an ABMS toolkit. The reactive agents complex biological systems may be explored, allowing unexpected represent bio-molecules and the logic for these agents is much relationships to be discovered, along with the testing of prior simpler than that of intelligent agents. The rules that depict hypotheses. It is hoped that this will reduce the subjectivity the goals of the bio-molecules, aim to produce generalisable inherent in the scientist’s choice of mathematical model, while at outcomes of the heterogeneous swarm. Although deliberately the same time allowing prior knowledge to help reducing statistical designing swarms to do specific cellular activities may sound instability in the modeling and mental overflow for the scientist. interesting and satisfying, it will be incapable of generating Results: The new explorative top-down approach for analysis generalised methods to capture all cellular activities. of massive time series data will be presented and illustrated Conclusion: The collective intelligence approach to modelling with some examples. The approach combines elements from intelligent cellular organisation is the ideal way of capturing several multivariate modeling traditions – bio-chemometrics, intelligent cellular level behaviour. This behaviour is fundamental to psychometrics and sensometrics - with various traditions within capturing the adaptive dynamics that occurs in a cell’s lifecycle. nonlinear dynamics, systems biology and computer science. For The most daunting task is to comprehend how biological illustration, various time series data sets, obtained by simulation structures emerge as an ongoing process of optimisation to from various mathematical models with known structures, were occupy functional niches. This optimisation is powered by taken as input time series data of “unknown” systems. The new various levels of selection in the biological hierarchy. Further this approach was able to reveal the underlying control structure approach could facilitate in capturing the mechanistic transition almost perfectly in the cases tested. between biological and pathological processes at the cellular level Conclusions: The proposed method opens new possibilities and assess the impact of various molecular species on cellular revealing regulatory structures in biological systems from the type Posters level activities. of empirical time series data that is expected to be available for Dedicated biologists in the near future. DS2-2-119 DS2-2-121 Fluctuations in model synthetic memories Jones, Nick Kinetic modeling of nucleotide excision repair Oxford University, Physics and Systems Biology, Oxford, United von Bornstaedt, Gesa1; Luijsterburg, Martijn S.2; van Driel, Roel2; Kingdom Höfer, Thomas1 1German Cancer Research Center, Heidelberg, Germany; Objective: To investigate simple synthetic memories relevant 2University of Amsterdam, Swammerdam institute for life for adaptive storage of external signals with the goals of both sciences, Amsterdam, Netherlands advancing our understanding of evolved biological signal processing and developing our own molecular components. Nucleotide Excision Repair (NER) is a multi-protein and multi-step Results: By analogy with the chemosensory adaptation repair system for DNA damage caused by UV light. In our model machinery of E. Coli, a memory element sitting inside an integral the main action is performed by five enzymatic reactions: the feedback loop is used to represent an extracellular signal, eg. initial and full unwinding of the DNA helix; double incision of the a chemoattractant concentration. The memory is noisy and damaged strand on both sides of the lesion; the re-synthesis of composed of uncoupled registers, but it has a simple multiscale the gap and finally ligation of the repaired DNA. structure so that its intracellular readout can efficiently encode Those five enzymatic steps separate six repair intermediates. information about extracellular quantities that vary by orders The first one represents the recognition of the damage, which of magnitude. A system with this component would be able to is performed by the strict sequential binding of two proteins. In distinguish between external signals pairs over a wide range the other repair intermediates located between the enzymatic without saturating its limited resources. Though an isolated steps, we assume random complex assembly in contrast to memory element is a poor representation of the external signal, sequential binding. The right complexes have to assemble, since an ensemble of such elements can be accurate and robust to certain protein complexes are needed to perform the enzymatic the intrinisc fluctuations of any particular element. If the integral reactions. For the repair intermediate of incised DNA not less than feedback on the system is turned off, or in the absence of seven individual diffusing components have to assemble to form extracellular signal, the variations in the elements can combine such multi-protein complexes. to show the non-Brownian fluctuations observed in the resting We assume that this formation is highly random and the proteins states of some natural systems. exchange rapidly, consistent with in vivo measurements of the Conclusions: A noisy, simple, biologically plausible, memory is exchange rates by FLIP (fluorescence loss in photobleaching). In presented which, using limited resources, can help encode, and our model, the observed long lasting steady state levels of some distinguish between, external signals which vary over orders of repair proteins are due to the time spend for having the right magnitude. complex assembly.

DS2-2-120

Elucidation of control structures by multivariate dynamic data-modelling Martens, Harald; Gjuvsland, Arne; Plahte, Erik; Omholt, Stig CIGENE, Ås, Norway

Objective: Modern biological measuring techniques are advancing rapidly, and there is a need for a corresponding

ICSB 2008 153 DS2-2-122 T-cell population. The high fidelity nature of the RT used ensures that the virions formed do not contain mutation. This aids the Functional criteria for evaluating metabolic models derived degradation of virions - both by available acting drugs and by from genome annotations natural immunity. Thus there is eventually a fall in the number of Gevorgyan, Albert; Fell, David; Poolman, Mark virions. Oxford Brookes University, School of Life Sciences, Oxford, United Kingdom DS2-2-124

Objective: The size of metabolic models derived from genome RNA fluctuations shape the dynamics of a negative annotations makes it laborious, and effectively impossible, feedback loop to identify errors and inconsistencies entirely by inspection. Rodriguez Martinez, Maria; Furman, Itay; Soriano, Jordi; Tlusty, During the construction of metabolic models of the bacteria Tsvi; Pilpel, Yitzhak Saccharopolyspora erythraea and Streptococcus agalactiae, and Weizmann Institute of Science, Rehovot, Israel the plant Arabidopsis thaliana, we sought to develop automated procedures to improve their functional quality and integrity. Objectives: Experimental measurements of gene expression Results: We have implemented the following checks: at the single cell level may differ markedly from measurements 1, Stoichiometric consistency. Further analysis of the taken across cell populations. An outstanding example is models requires that mass is conserved, but we can locate provided by the p53-mdm2 negative feedback loop, where single inconsistencies in a network even when the reactions taken cell experiments of both p53 and mdm2 expression show a separately appear mass balanced. series of sustained anti-correlated pulses, while cell population 2. Unemployed enzymes. These are usually enzymes that appear experiments show damped oscillations. Mathematical modeling in the annotation, but none of the reactions they catalyze can of single cell dynamics that would reproduce the measured have a non-zero rate in any steady state of the model. Some of dynamics remains a challenge. the potential reasons for this require experimental investigation, Opposed to some previous models, in this work we build a fully but one possibility can be investigated as follows. stochastic model of such a negative feedback loop that explicitly 3. Ambiguous annotations. Annotations are often based on takes into account the RNA kinetics. In our framework we can Dedicated some form of sequence matching, with the best match selected. investigate the role of RNA stochastic fluctuations as a leading Posters Sometimes (e.g. with PRIAM), more than one match can be force to produce a sustained excitatory behavior in single cells. made, and the other matches may still be significant. We test Results: Our explicit modeling of the kinetics of both transcription whether the number of unemployed enzymes can be decreased and translation allows the dissection of the contribution of each by exchanging the best match for one of the other possibilities. In step towards the final protein noise level. We find that RNA this way we generate testable hypotheses about changes to the random fluctuations can lead to pulses of protein expression. The annotation. pulses disappear when the kinetics or the stochastic nature of Conclusions: To the extent that we have tested published mRNA is ignored. A deterministic study of this network fails both genome-scale metabolic models, we find that stoichiometric to find an oscillatory pattern or to predict the average protein inconsistency is not a major problem, but the presence of concentration. unemployed enzymes and reactions is common. We recommend Motivated by the recent realization of the importance of non- that they should be identified and removed from the count of coding RNAs in transcription regulation, we also include an the functionally active enzymes and reactions. In the case of our antisense RNA transcript that binds to the mRNA and blocks model of S. agalactiae, we have proposed new annotations for translation. We find that the pulses are partially smoothed out a number of genes, which are being tested by our experimental under sufficiently strong mRNA and antisense interaction. collaborators through gene knockouts and transcriptional Conclusions: Our modelling study emphasizes the utilization of correlations. stochastic computation methods as an engine to reproduce the dynamics observed in single cells. DS2-2-123 We show that, opposed to common intuition, the stochastic dynamics of RNAs can be amplified by a negative feedback loop, Drug effectiveness enhancement in T-cell populations by and produce sustainable oscillations of protein expression. Such modified reverse transcriptase - an in silico study dynamics cannot be obtained using a deterministic approach. Guda, Sanjay Kumar; Badmi, Kalyan Venkatesh; N, Hemanth Antisense RNA has the ability to filter out mRNA fluctuations to Kumar; K.R., Vivek Sagar some extent, thus reducing the noise in single cells and bringing PES Institute of Technology, Biotechnology, Bangalore, India a cell population closer to the state expected from a deterministic model. Objective: The current emphasis in treatment of AIDS is on enhancing the mutation of HIV to such an extent that it loses DS2-2-125 potential for replication. In this work we propose a method whereby the major hindrance to the treatment of AIDS, the high Definition and application of Mauritius maps to analyze mutability of the HIV, is reduced or stopped. This is critical to allow biochemical systems available drugs to continue their work and avoid the development Koch, Ina1; Ackermann, Joerg2 of drug resistant strains. A drug’s applicability today is for few 1Technical University of Applied Sciences Berlin / Max Planck cycles only (4 to 6). Ensuring that the pro-virus formed is a clone Institute for Molecular Genetics, FB VI / Dept. Computational of the infecting strain of HIV will enable drugs to continue to act Molecular Biology, Berlin, Germany; 2fluIT Biosystems, St. on the virions generated and thus significantly extend the number Augustin,, Germany of cycles for which they are effective. A chemical engineering kinetics type model was built using Cell Designer 4.0. The model Objectives: Because of experimental limitations or ethical was used for studying effect of a modified reverse transcriptase reasons, often more qualitative than quantitative data is known. (RT) and two anti-retro viral drugs. The effect of this approach on To estimate the network behavior qualitative concepts have been the populations of infected T-cells is also modeled. developed and applied, for example, elementary mode analysis Results: An RT molecule with high fidelity was designed. This for metabolic networks or invariant analysis for biochemical Petri ensures that more clones of the parent are formed. This RT was nets. Both can be interpreted as biological functional modules delivered to the cell at high concentrations, thereby increasing or pathways. With increasing size and complexity of networks, the probability of its packaging. The Km of native RT is 0.12 and the number of elementary modes and T-invariants, respectively, that of the novel therapeutic is increased around 10 folds and the growths exponentially which makes their detailed exploration factor by which it gets packaged is also increased by 7 fold. difficult. To facilitate the investigation of functional dependencies Conclusions: The population dynamics study showed that there between pathways we have developed a new graphical is an initial increase in the virions formation across the entire representation we termed Mauritius maps, which can be applied

154 ICSB 2008 to both, elementary modes and T-invariants, to efficiently perform, DS2-2-128 for example, in-silico knockout analyses. Results: Mauritius maps represent labeled binary trees, where A transcriptional regulatory switch underlying B-cell vertices denote transitions, describing chemical reactions. The terminal differentiation and its disruption by dioxin root of the tree is located in the lower left corner. Horizontal edges Bhattacharya, Sudin1; Conolly, Rory B.2; Andersen, Melvin E.1; connect transitions belonging to the same T-invariant. Vertical Kaminski, Norbert E.3; Zhang, Qiang1 edges indicate a new T-invariant that consists of the transitions 1The Hamner Institutes for Health Sciences, Division of in the left subtree and the transitions in the new branch. Starting Computational Biology, Research Triangle Park, United States; with the root, the horizontal line is labeled by the most important 2U.S. Environmental Protection Agency, National Center for reactions, i.e. by those reactions exhibiting the highest impact, Computational Toxicology, Research Triangle Park, United States; if they would be knocked out. For example, a reaction, which 3Michigan State University, Department of Pharmacology and is a member of all T-invariants, can be considered to be “most Toxicology, East Lansing, United States important”. Through a knockout of a reaction, one part of the net remains active, while another part looses its biological function. The terminal differentiation of B cells in lymphoid organs into The knockout of a single transition affects all pathways described antibody-secreting plasma cells upon antigen stimulation is a by the corresponding right child and its successors. Hence crucial step in the humoral immune response. The architecture the branch of the right child denotes the part of the net which of the B-cell transcriptional regulatory network consists of looses its biological activity. The branch of the left child remains coupled mutually-repressive feedback loops involving the three unaffected by the knockout. transcription factors Bcl6, Blimp1 and Pax5. This structure forms Conclusions: Mauritius maps serve as ordered representation of the basis of an irreversible bistable switch directing the B-cell sub-pathways and their functional dependencies, in particular, for to plasma cell differentiation process – i.e., the switch remains a huge number of T-invariants. Mauritius maps facilitate in-silico on even after the activating stimulus (antigen) is removed. The knockout experiments avoiding new computational effort. environmental contaminant 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) is known to suppress the humoral immune response by DS2-2-127 interfering with this differentiation program. We have developed a computational model of the biochemical pathways that Complexity-based modelling of stem cell fate decisions: regulate B-cell differentiation, and the molecular mechanism by Mesoscale analysis of regulation using microarray data which TCDD impairs this process through the action of the aryl Burden, Frank; Halley, Julianne; Winkler, Dave hydrocarbon receptor (AhR). Using a kinetic model and bifurcation CSIRO, CSIROMolecular and Health Technologies, Private Bag analysis, we propose that TCDD regulates the proportion of

10,Clayton South, Australia B-cells that differentiate into plasma cells by raising the threshold Posters dose of antigen lipopolysaccharide (LPS) required to trigger the Dedicated Objective: In the context of stem cell decision-making, the differentiation switch. We also show that stochastic modeling of generation of diverse responses equates to the priming of gene expression, which allows cell-to-cell differences in content lineage-specific modules of genes. This idea is consistent with of signaling proteins, introduces distributional characteristics to the multi-lineage priming hypothesis. Priming can be triggered the timing and probability of differentiation among a population by stochastic gene expression, initiating local avalanches of of B-cells. This cell-to-cell variability is likely to be a key gene expression, drawing gene modules closer to their critical determinant of dose-response and sensitivity of individual cells to (decision-making) points. differentiation. Results: Use of microarrays to follow cell lineages poses an This presentation may not reflect the official policies of the U.S. interesting and important class of mathematical problems, those EPA. in which the number of variables greatly outweighs the number of observations. We have used two approaches to carry out DS2-2-129 this task using microarray data sets from cell lineages. We use a sparse Bayesian feature reduction method based on an evidence In silico modelling of cation homeostasis in maximisation algorithm [Figueiredo] to identify a small set of Saccharomyces cerevisiae relevant genes. We have also developed method for mesoscale Gerber, Susanne1; Borger, Simon1; Lichtenberg-Fraté, Hella2; network modelling of gene regulation that employs a new, Ludwig, Jost2; Shabala, Sergey3; Shabala, Svetlana3; Kschischo, efficient, linear algebra approach (MesoNet), originally inspired Maik4; Kahm, Matthias4; Klipp, Edda1 by a recursive network model [Geard and Wiles]. Both of these 1MPI for Molecular Genetics, Computational Systems Biology methods identify a few genes that can be used to predict the Group, Berlin, Germany; 2University of Bonn, IZMB, Bonn, expression of many thousands of genes during erythropoiesis and Germany; 3Universityof Tasmania, School of Agricultural Science, embryogenesis. Tasmania, Australia; 4University of Applied Sciences Koblenz, Conclusions: The genes identified by the EM method, and Biomathematics, Remagen, Germany those highly correlated with them, were able to predict lineage specification and were relevant to the developmental pathway. Objective: Homeostasis of cations and maintenance of The MesoNet approach identified a few key regulatory genes in membrane potential are essential for survival of cells, for instance a sequence of microarray data that were able to recapitulate the for nutrient uptake. Increased cellular concentrations of ions expression of the majority of the genes. This talk will describes generally are toxic and hyperpolarisation of the membrane causes these methods and illustrates them using murine data from the non-specific influx of ions. In two ways ions are transported literature [Welch, Hamatani]. across cell membranes: passively by diffusion or actively by M A.T.Figueiredo, IEEE transactions on pattern analysis and transmembrane proteins. In the yeast S. cerevisiae potassium and machine intelligence, - PAMI, 2003,25, 1150-1159. hydrogen mainly contribute to membrane potential build up. The N.Geard,J.Wiles, Artificial Life, 2005, 11, 249-268. transmembrane ion fluxes are coupled. For instance sodium plays T.Hamatani, M.G.Carter, A.A.Sharov, M.S.H.Ko, Developmental a role in the regulation of intracellular potassium. We provide a Cell, 2004,6,117-131. non-equilibrium thermodynamic description of cation homeostasis J. J.Welch, J.A.Watts, C.R.Vakoc,Yu Yao,Hao that accounts for the coupling between the ion fluxes. Wang,R.C.Hardison,G.A.Blobel, L.A.Chodosh, and M..J.Weiss, Results: Data for ion fluxes, ion concentrations and membrane Blood, 2004, 104,3136-47. potential turn the thermodnynamic description of cation homeostasis into a linear equation system. The coefficients of a symmetric square matrix were determined. The number of coefficients is n(n+1)/2 where n is the number of ion species. Positive coefficients entail correlation, negative coefficients anti- correlation between forces and fluxes. Conclusions: Membrane potential and concentration imbalances

ICSB 2008 155 cause constant ionic fluxes across the cell membrane. The fluxes Results: Among β-TrCP isoforms, interestingly, β-TrCP1 of the different species respond to each other in order to achieve is stabilized by β-catenin/TCF activation thereby mediates cation homeostasis. A description of homeostasis therefore has a negative feedback loop for β-catenin, whereas β-TrCP2 to include these interdependencies in order to achieve its goal. stabilization is suppressed by β-catenin/TCF activation leading to The appropriate thermodynamic approach is presented here. a positive feedback loop for β-catenin. Moreover, this difference Furthermore, the validity of the model in view of ionic flux data in their stabilization results in an incoherent feed-forward loop and the so fare known kinetic parameters including consecutive between β-catenin/TCF and IκBα. From our mathematical model future working steps will be discussed. Future work will seek to and in-silico simulations, we found that this incoherent feed- give a mechanistic description of the couplings between the ionic forward loop through β-TrCP1 and β-TrCP2 causes a biphasic species. behavior of NF-κB activity for Wnt stimulus in a certain range of the stabilization rate. In addition, we revealed that aberrant up- DS2-2-130 regulation of β-catenin under an abnormal condition such as APC mutation induces over-expression of NF-κB which is responsible Modeling DNA replication in Saccharomyces cerevisiae for anti-apoptosis and cell transformation. Spiesser, Thomas; Klipp, Edda; Barberis, Matteo Conclusions: Throughout this study, we have unraveled that Max Planck Institute for Molecular Genetics, OWL - Otto Warburg different roles of β-TrCP1 and β-TrCP2 give rise to the biphasic Laboratories, Berlin, Germany response of NF-κB to Wnt stimulus under a normal condition and induces a high over-expression of NF-κB by destroying the Objective: In eukaryotes DNA replication is considered incoherent feedforward structure at a tumorigenic stage. to proceed according to a precise program in which each chromosomal region is duplicated in a defined temporal order1. DS2-2-132 Recent studies however, revealed an intrinsic temporal disorder in the replication of yeast chromosome VI2. Here we provide a Engineering-genes based modelling of biochemical model of the DNA replication dynamics to study the temporal reaction networks sequence of origin activation in the budding yeast. The model Gormley, Padhraig; Li, Kang is composed out of four parameters that influence the DNA Queen’s University Belfast, Belfast, United Kingdom Dedicated replication system: the lengths of the chromosomes3, the explicit Posters chromosomal positions6 for all replication origins as well as Objective:One of the goals of systems biology is to be able to their distinct initiation times6 and the replication fork migration model the activity and interactions of molecular species within rate1,4,5. The simulations provide new insights into the complex the cell. However due to the sheer complexity of the sub-cellular system of DNA replication, enabling us to reinterpret the existing environment it is difficult to determine which protein species experimental data in the light of the above defined parameters. interact with each other in the form of networks of reactions and Results: The constructed model for the DNA replication is able consequently, which gene products affect which gene targets. to reproduce experimental data in form of replication profiles. The objective therefore is to automatically infer these reaction Furthermore, simulations show replication kinetic patterns, networks solely from measurement data in order to create system which match experimental ones. The dynamics of the replication models in the form of coupled differential equations describing process was monitored during simulations of the randomly the concentration levels of molecular species over time. This perturbed replication system. Hereby, severe loss of origin is investigated here using a co-evolutionary algorithm that function (reduction to 50%) showed only little influence on automatically reverse engineers the structure and parameters the replication dynamics. Moreover, deletions of origins were of differential equation models from sets of time series data. simulated systematically, providing predictions to be tested Model structures are selected from a pool of fundamental terms experimentally. extracted from a priori knowledge of the underlying biochemical Conclusions: By constructing a spatiotemporal model on process. These fundamental nonlinear terms are known as the basis of four replication system parameters, and using but ‘Engineering-genes’ in the control engineering literature and measured values throughout the whole analysis, we were able essentially help to focus the search on biologically plausible to provide an accurate model for the DNA replication dynamics, models, improving speed and accuracy of convergence. which can be used to study the spatiotemporal order and Results: The method was applied to several simple low explains trends exhibited during the DNA replication process. dimensional systems that have biological relevance, such as 1. Raghuraman et al (2001) Science 294 (5540) 115-121. the example below, derived from mass action kinetics and with 2. Czajkowsky et al (2008) J Mol Biol 375 (1) 12-19. fractional kinetic orders. The best model inferred using our 3. Pruitt et al (2007) Nucleic Acids Res 35 D61-D65. method and the conventional evolutionary approach is given after 4. Rivin et al (1980) J Cell Biol 85 (1) 108-115. 20 runs in each case. 5. Yabuki et al (2002) Genes Cells 7 (8) 781-789. Target system: 0.5 6. Nieduszynski et al (2007) Nucleic Acids Res 35 D40-D46. dx1/dt=2x2 0.5 dx2/dt=x2x1 DS2-2-131 Eng-genes method: 0.4909 dx1/dt=2x2 0.6541 Cross-talks between NF-κB and Wnt/β-catenin pathways dx2/dt=0.89x2x1 and their implications in the regulation of tumorigenesis Conventional method: 1 2 1 0.4772 Shin, Dongkwan ; Lee, Hyeon-Woo ; Cho, Kwang-Hyun dx1/dt=1.8558x2 -0.23x1 1 1.335 0.5865 Korea Advanced Institute of Science and Technology(KAIST), dx2/dt=1.5639x2 x1 -1+2.092x2 Department of Bio and Brain Engineering, Daejeon, Republic Conclusions:The results show that compared to the of Korea; 2Kyung Hee University, Department of Pharmacology, conventional method, our eng-genes technique of biasing the Seoul, Republic of Korea term-pool with a priori knowledge of the underlying biochemical reactions improves the convergence of the reverse engineering Objective: There are some recent reports on the cross-talks process, resulting in more accurate and transparent models. Also through beta-transduction repeats-containing proteins (β-TrCP) by using the algorithm’s partitioning feature to decouple system between the Wnt/β-catenin and NF-κB pathways that are equations and reduce the dimensionality, it has potential for involved in malignant transformation of many types of human inferring higher dimensional systems provided that large enough cells. These ubiquitin ligases target both β-catenin and IκBα sets of real data are available. via the ubiquitination and subsequent degradation. An aberrant overexpression of β-TrCP is often found in several cancer cell lines, indicating that β-TrCP isoforms between the Wnt/β-catenin and NF-κB pathways are expected to play some role in the regulation of tumorigenesis.

156 ICSB 2008 DS2-2-133 primary optimization target. Here, these constraints are put together to form a modeling framework. A logic-based approach for modeling genotype-phenotype Conclusions: We are confronted with the challenge to find a relations in Campylobacter trade-off: on one hand a metabolic model of a higher cell must Tamaddoni-Nezhad, Alireza1; Barton, Richard2; Hitchen, Paul3; be sophisticated enough to cope with its high complexity in their Kay, Emily4; Lesk, Victor3; Turner, Frances3; Dell, Anne3; Rawlings, functions and their demands, on the other hand the number of Christopher1; Sternberg, Michael3; Wren, Brendan4; Muggleton, model parameters must be small enough to be supported by Stephen1 restricted experimental data and it must be simple enough that 1Imperial College London, Dept. of Computing, London, United it allows to draw concise qualitative biochemical knowledge Kingdom; 2Imperial College London, Division of Molecular from it. On top of that, the computation efforts should not Biosciences, London, United Kingdom; 3Imperial College London, impede screening of scenarios and gradual changes of cellular Division of Molecular Biosciences, London, United Kingdom; constraints. At the present state of information on higher cells 4London School of Hygiene and Tropical Medicine, Dept. of constrained flux-balance models give the most realistic solution. Infectious and Tropical Diseases, London, United Kingdom DS2-2-135 Objective: A central challenge in computational systems biology is to build models that will allow us to “read across” between Varying the level of abstraction in stochastic molecular different ‘omics datasets and to use diverse biological prior dynamics models knowledge. Here we highlight an exemplar research programme Mura, Ivan; Palmisano, Alida; Sedwards, Sean; Dematte’, at the Centre for Integrative Systems Biology at Imperial College Lorenzo; Romanel, Alessandro (CISBIC). We present a logic-based approach to integrative Centre for Computational and Systems Biology, Trento, Italy systems biology and demonstrate this by building predictive models of genotype-phenotype relations in Campylobacter jejuni. Objective: Systems biology calls for models at different Results: Abductive logic programming is used to infer abstraction levels: elementary biochemical reactions, interactions hypotheses about the function of genes from observed changes among macro-components such as genes and polymerase, in glycans structures (mutant data) together with background intercellular signaling mechanisms, competition of predators and knowledge which includes known metabolic networks, glycan prey are some examples. Systems can be modeled at different structures and enzyme/gene functions. This background abstraction levels, depending on modeling objectives, amount/ knowledge is extracted from biological databases including KEGG quality of available data, trade-offs between computation cost/ and converted into Prolog. Experiments on Lipo-Oligo-Saccharide detail. Effective modeling tools have to support variable levels

(LOS) synthesis pathway and relevant mutants data suggest of abstraction. One current limitation of stochastic modeling Posters that the model can be used to predict (with predictive accuracy approaches based on Gillespie’s algorithm is that they deal with Dedicated more than 95%) the function of a knocked-out gene even if part the elementary levels of biochemical reactions only. We aim at (up to 28%) of background knowledge on reactions and gene extending stochastic modeling formalisms to allow a flexible functions are removed from the model. In this setting abduction is choice of abstraction level. Results: We propose to extend used to infer the function of genes and also to fill the gaps in the propensity functions defined as per Gillespie approach towards a pathways. more general formulation, which we call rate functions. Propensity Conclusions: A logic-based representation & inference (i.e. functions only express the speed of reactions by product of rate Abductive ILP) has been used to infer predictive models of constants and number of simultaneous reactions taking place. genotype-phenotype relations in Campylobacter from mutants Rate functions allow defining more complex mathematical forms data and a background knowledge which can potentially be of rates and logical conditions on the occurrence of events in the incomplete. However, initial results suggest that the present models. Michaelis-Menten, Hill as well as any function providing model can be only used in cases where the pathways are an accurate description of the rate at which biological phenomena relatively complete (e.g. more than 70%). On going research aims proceed and of the conditions that regulate biochemical at extending the model in order to use additional empirical data transformations, can be directly transferred into rate functions. such as metabolites and gene expressions profiles (metabolome The practical issues of application of this approach were studied and transcriptome data) as well as additional background with Stochastic Petri Nets, and an implementation realized for knowledge such as protein-protein interaction networks. the BlenX and Cyto-Sim tools. Gene transcription regulation and yeast cell cycle were modeled with the extended formalism DS2-2-134 supporting rate functions. Conclusions: The extension of modeling tools to support Constrained optimization sheds light in the metabolic general rate functions removes the constraint of elementariness of functioning of higher cells reactions and allows including highly expressive modeling features Hoppe, Andreas; Hoffmann, Sabrina; Holzhuetter, Hermann- in Gillespie based stochastic approaches. As a corollary, this Georg extension also allows a natural mapping between deterministic Charité, Hochschulmedizin Berlin, Institute for Biochemistry, and stochastic models of biological systems. Berlin, Germany DS2-2-136 Objective: To model single-cell organisms by an optimization approach a simple target function suffices since these cells only Using the photosynthetic apparatus of rhodobacter need to grow and replicate. However, higher organism cells sphaeriodes as a testcase for a comprehensive systemic such as the human hepatocyte have a multitude of different reconstruction of a metabolic system functions: self maintenance, substrate supply to the organism, Geyer, Tihamer; Mol, Xavier; Blass, Sarah; Lauck, Florian; Helms, transformation of waste products, clearance of toxic substances, Volkhard homeostasis etc. . Thus, the effectivity of such a cell can only be Saarland University, Center for Bioinformatics, Saarbrücken, measured in the context of the whole organism. Germany Results: New constraints for FBA (flux-balance analysis) raised their applicability for higher cells: 1. Restricting net flux directions Objective: For simulations of metabolic systems, often only a by thermodynamic laws lessens the necessity to fix these part of the required rate constants and parameters are known. directions heuristically, thus, increasing the potential diversity of We investigated, how these can be determined consistently with a metabolic functions a model can describe. 2. The consideration systemic model setup which incorporates all available information of a (non-perfect) fitness of the cell allows to model situations about the system and a simultaneous parameter fitting against a where the cell is confronted with a too heavy load or shortage number of different dynamical experiments. For this process we of selected substrates. 3. A majority of cell’s energy is devoted used the simple and well characterized photosynthetic apparatus to protein synthesis, thus, enzyme costs are considered as the of the purple bacterium Rhodobacter sphaeroides.

ICSB 2008 157 Results: During the setup process, the size of the chromatophore DS2-2-138 vesicles housing the photosynthetic apparatus put restrictions on, e.g., the relative placement and stoichiometries of the proteins An optimization based approach to uncover metabolic and on their connectivity. Vice versa, kinetic constraints aided objectives pursued by changes of enzyme levels in the spatial reconstruction of the very small chromatophores, Hoffmann, Sabrina; Holzhuetter, Hermann-Georg which carry a total of less than a hundred membrane proteins. Charité - Medical Faculty Berlin, Institut of Biochemistry, Berlin, To account for the fluctuations in this small system, a molecular Germany stochastic simulation was set up, where each protein with their explicit binding sites and internal states is modeled individually. Objective: Expression profiling and proteomic techniques In this bottom-up approach, the model for each of the proteins reveal significant variations in the level of thousands of mRNAs is constructed from the relatively well understood microscopic and proteins upon environmental changes such as substrate charge transfer or association and dissociation reactions, while depletion, oxidative stress or hormonal stimulation. However, the overall connectivity is not fixed a priori, but emerges from in most cases the functional implications of these variations the dynamic interplay of the different proteins. We then used remain elusive. One crucial problem complicating their functional an evolutionary algorithm to fit the rate constants and system interpretation of is that changes of protein levels do not simply parameters such that a set of eight experiments, which range translate into equivalent changes in the rate of the associated from quasi steady state situations to fast multi-flash-triggered chemical processes due to many regulatory mechanisms scenarios, is reproduced best. operating below the gene expression level. Here, we outline a Conclusion: The chosen approach with a stochastic simulation theoretical concept to exploit information on (relative) changes in together with an evolutionary search for sets of model parameters the level of metabolic enzymes for the prediction of (relative) flux is a promising route to a systemic modelling and understanding of changes in the underlying metabolic network. metabolic systems. These techniques can also be applied to gene Results: The prediction method comprises two main steps. regulation or mixed systems of metabolism and regulation. We First, we approximate (unknown) flux changes by a linear also discuss possible extensions of the current model of bacterial combination of so-called minimal flux modes (MFM) each photosynthesis. representing a specific flux distribution just required to accomplish the production of only one of the numerous functionally relevant Dedicated DS2-2-137 output metabolites. Second, the unknown coefficients of this Posters decomposition are chosen such that a maximal correlation with Glial response to enhanced neural activity: Feedback and observed differential expression data is obtained. Based on spatial buffering simulated enzyme level scenarios in a metabolic model of the Øyehaug, Leiv; Østby, Ivar; Einevoll, Gaute T,; Omholt, Stig W. erythrocyte we demonstrate the predictive capacity of our method Norwegian University of Life Sciences, Centre for Integrative on the established flux distribution, but also on the change of Genetics, CIGENE, Ås, Norway output metabolite synthesis that necessitated the observed change in enzyme levels. Objective: Neuronal activity triggers release of neurotransmitters Conclusions: For this small and well studied model system and ions from the neuron. In order to investigate the response promising results have been obtained, which motivate further of glial cells to enhanced neuronal activity, we developed two extension of the concept. In future, the approach will also be dynamic models: The first one consists of a description of applied to changes of mRNA levels instead of changes in enzyme important glial membrane processes and a Hodgkin-Huxley levels. Here, the challenge remains to understand the non- type model of neuronal membrane processes. The second one obvious relationship of mRNA profiles to protein levels. Thus, includes intra- and extracellular ion transport along a spatial axis the presented method can be taken as a first step towards the and can be used to investigate conditions for spatial buffering, a interpretation of gene expression data with respect to metabolic process by which potassium is taken up (by glia) at sites where, activity. due to neuronal activity, the potassium concentration is high, and instantaneously released at distant sites where the potassium DS2-2-139 level is normal. Results: Neuronal activity, mimicked by applying a current to Statistical methods for detecting functional abundance in the neuronal membrane, causes a rise in the level of extracellular metagenomes + + 1 2 3 potassium, [K ]o. Clearance of K is highly dependent on the Dalevi, Daniel ; Kristiansson, Erik ; HugenHoltz, Phil ; Ivanova, parameters associated with the Na+/K+/2Cl- (NKCC1) and Natalia4; Markowitz, Victor5 + - 1 2 Na /2HCO3 (NBC) cotransporters. For weak cotransporter and Dept. Computer Science, Göteborg, Sweden; University of + 3 for suitable choices of parameter values, the high [K ]o may cause Gothenburg, Dept. Zoology, Göteborg, Sweden; Joint Genome self-sustained neuronal activity, i.e. seizures. Institute, Microbial Ecology Program, Walnut Creek, United Using the spatial model, we show that the choice of certain model States; 4Joint Genome Institute, Genome Biology Program, parameters decides whether the astrocyte will provide a dynamic Walnut Creek, United States; 5Lawrence Berkeley National influx or a constitutive efflux of potassium through the potassium Laboratory, 1Biological Data Management and Technology + (Kir) channels during increased [K ]o in the ECS. Indeed, the effect Center, Berkeley, United States of spatial buffering depends on the pumping rate of the Na/K/ ATPase pump, conductance properties of the ion channels in the Objective The goal of metagenomics is to determine and analyze membrane, and the potassium diffusion coefficient. the conjoined genome of all the organisms in a community or an Conclusions: The model results indicate that the astrocytic environment. There are presently many studies publically available NKCC1 and NBC cotransporters contribute to the clearance including data from seawater, soil and human gut. What is of extracellular K+ and to maintaining a sound extracellular common to these samples is that they are highly fragmented and environment. The spatial buffering mechanism, acting in concert contains only a small sample of the entire metagenome. This is a with the Na/K/ATPase pump and voltage gated channels, is consequence of non-uniform species abundances and the costly dependent on the properties of these mechanisms and on the process of generating the data. Statistical methods are therefore potassium diffusion in order to be efficient. necessary to detect relative differences in species or functional frequencies. A common problem is to test whether a gene- family (e.g. COG, TIGRFAM etc), or a set of gene-families (e.g. pathways), is more abundant in one sample compared to another. Results We compare tests for identifying differences of individual gene-families between samples of metagenomes. Several methods for analyzing sets of gene-families are also proposed; including a Gaussian-sum statistic, Fisher enrichment analysis and a Poisson-regression model. Simulated metagenomic

158 ICSB 2008 data was generated in a realistic setting from observed gene- transduction networks, their spatial-and-temporal dynamics family distributions of isolate genomes and non-uniform species features are analyzed. Through the simulation of Michaelis-Menten distribution. All procedures are benchmarked and evaluated using kinetics of the related signal pathways, the movement direction of both the simulated data and some real metagenomic samples. the molecular motors that is with the maximum probability value Conclusions We show that a binomial test performs well for is estimated in mathematics. This result provides us a measure to identifying over-abundant gene-families. Furthermore, for sets of select the optimal pathway that activates the motor proteins. gene-families we find that the Poisson-regression model performs best, closely followed by the Gaussian-sum statistic. The tests DS2-2-142 were applied to samples from microbiomes of human gut and a sunken whale skeleton carcass to confirm the underlying model Characterization of mutants by Fourier analysis of assumptions. stochastic simulations of yeast cell cycle Sedwards, Sean DS2-2-140 MSR -UNITN CoSBi, Trento, Italy

Comparative analysis of logical models of the cell cycle in Objective: Stochastic simulation of the Novak & Tyson (2002) cell eucaryotes cycle model has proven to have greater discriminatory power in Faure, Adrien1; Naldi, Aurelien1; Chaouiya, Claudine1; Ciliberto, revealing yeast mutants than the traditional approach using ODEs Andrea2; Thieffry, Denis1 (Csikasz-Nagy & Mura 2008). Rudimentary statistical measures 1TAGC - INSERM U928, Marseille, France; 2IFOM, Milano, Italy of the fundamental period of oscillation of such simulations have shown good accord with experimental data, while visual Objective: The network controlling the cell cycle is highly inspection of the time courses of concentrations clearly indicate conserved among eucaryotes. Using GINsim (http://gin.univ-mrs. anomalies hidden by deterministic simulations. We have now fr/GINsim), we develop logical models for the regulatory networks applied Fourier analysis to multiple time courses of various controlling cell cycle variants in various organisms. These models mutants in order to formalize the discriminatory process. are then used for comparative topological and functional analyses. Results: Fourier analysis of deterministic simulations produce Results: Based on literature data as well as on homology a discrete spectrum which characterizes the time course relationships with components of a generic mammalian cell in the frequency domain. Being deterministic, repeated cycle model (Fauré et al, 2006), we have delineated a model simulations produce identical spectra and say no more than accounting for three variants of the cell cycle in Drosophila: early is available in a single run. By contrast, repeated stochastic embryonic syncytial cycles, endocycles and canonical cycles simulations reveal additional information about the model, such

(Swanhart et al, 2005). In parallel, we are developing a model of that the accumulated Fourier spectrum takes on the nature Posters Arabidopsis endocycles. Comparing these models with those for of a continuous distribution. The shape of this distribution Dedicated the mammalian and budding yeast cell cycles reveals that, in each characterizes the distribution of simulations and thus, in our case, case, a negative circuit involving Clb2 and Cdc20’s homologs, as the yeast mutants. In this way we are able to give a compact well as several short, two-elements positive circuits involving Clb2 formal description of observed behaviour. and one of its regulators are playing crucial dynamical roles. Such Conclusions: Stochastic simulation reveals additional information combinations of positive and negative circuits have recently been about a model but such information needs a formal interpretation associated with robustness and tunability of biological oscillations to reduce the apparent complexity. Fourier analysis is a natural (Tsai et al, 2008). choice for the interpretation of oscillatory data and therefore Furthermore, the precise sequence of activations and inhibitions finds good application in this context with cell cycle mutant of the different components of the cell cycle regulatory network is characterization. The combination of stochastic simulation plus only marginally affected by fine tuning of time delays or priorities Fourier analysis is a powerful tool that has wider application in associated with these components, here again indicating robust both oscillatory systems and those with more general transient control of cell cycle progression (see also Li et al, 2004). behaviour. Conclusions: The comparison between logical models of cell DS2-2-143 cycle networks enables the delineation of conserved functional features beyond component homologies, in particular the crucial Dynamical analysis of hb border formation under variable roles of specific regulatory circuits. Furthermore, the recurrent Bcd morphogen in a model of the drosophila gap gene robustness we observe substantiates an incremental modelling network strategy, starting with simplified Boolean synchronous models Gursky, Vitaly1; Manu2; Surkova, Svetlana3; Samsonov, Alexander1; to progressively define more elaborate discrete models by Samsonova, Maria3; Reinitz, John2 considering multilevel components or kinetic refinements. 1Ioffe Physico-Technical Institute, Theoretical Department, St. References: Fauré et al (2006). Bioinformatics 22: e124-31. Li Petersburg, Russian Federation; 2Stony Brook University, Stony et al (2004). Proc Natl Acad Sci USA 101: 4781-6. Swanhart et al Brook, NY, United States; 3St. Petersburg State Polytechnical (2005). Methods Mol Biol 296: 69-94. Tsai et al (2008). Science University, St. Petersburg, Russian Federation 321: 126-9. Objective: In order to better understand processes of robustness DS2-2-141 and canalization (error correction) in early embryos, we analyze the dynamical effects of varying spatial profiles of Bicoid (Bcd) A dynamical network model inspired by molecular motors on the formation of a border of expression of the Drosophila gap Liu, Jian-Qin; Oiwa, Kazuhiro gene hunchback (hb) in blastoderm stage embryos. National Institute of Information and Communications Technology Results: We analyzed attractors and their basins of attraction (NICT), Kobe Advanced ICT Research Center (KARC), Kobe, in a dynamical model parameterized by position on the anterior- Japan posterior (A-P) axis. This model, which is well supported by extensive experimental data, was tested against a family of over Based on the biophysics mechanism of molecular motors, 80 Bcd gradients obtained from individual embryos. Here we modeling molecular movement in cells is important for us to study show that the formation of the hb border is associated with the the dynamics of genetic processes happening in cells, where the intersection of the spatial gradient of maternal Hb protein and a movement of motor proteins can be activated by the signaling boundary between the basins of two different attractors. Although pathways. By integrating the methodology of systems biology and the precise identities of these attractors may vary, their effect on synthetic biology, a dynamical network model is proposed for a the Bcd dependence of the hb border is remarkably uniform and mesoscopic description of the signaling mechanism of temporal is a direct consequence of the geometrical properties of attraction dynamics in signal transduction pathway networks and a spatial basin boundaries. dynamics of motor proteins. Through the stochastic measurement Conclusions: The dynamical analysis reveals high stability of of the movement of molecular motors and the simulation of signal hb border positioning in response to varying Bcd concentration.

ICSB 2008 159 The observed robustness is a consequence of two mechanisms. time series data by global gene expressions with focusing on First, the initial conditions and basin boundary have opposite extreme values. dependencies on A-P position. Second, there is a compensating Results: We analyze Affymetrix array data for various effect by Bcd itself on attraction basin boundaries which reduces experimental conditions of human cells retrieved from the variation for very low or high levels of Bcd. ArrayExpress database. The characteristic of our analysis is that To our knowledge, this is the first study of canalization (error we focus on minimum and maximum values of scaled time series correction) in a population of individuals in which the dynamical data divided by their initial values, reflecting on the most activated effects of variation of a quantitative embryonic determinant are and inhibited situations among the observation points. We then precisely analyzed. show a non-trivial peak in both minimum and inverse of maximum The study was supported by NIH Grant RR07801, CRDF GAP distributions at similar positions. The two peaks are rather robust award RUB1-1578, RFBR-NWO grant 047.011.2004.013, RFBR under different experimental conditions and different normalization grants 08-01-00315a and 08-04-00712a. procedures between MAS5 and RMA. Conclusions: The non-trivial peak phenomenon can be DS2-2-144 statistically derived from microarray data of human cells by paying attention to the extreme expressed levels on each gene. We Symmetrical aspect in intracellular global dynamics of presume that two peaks are generated from an aggregate of gene mRNA expression feedbacks. Namely, each of peaks in minimums and Asakawa, Takeshi1; Sazuka, Naoya1; Nagashima, Takeshi2; maximums is a quantitative threshold of feedback effect of up and Mariko, Hatakeyama2; Shiraishi, Tetsuya3 down regulated genes respectively. The interaction between gene 1Sony Corporation, Tokyo, Japan; 2RIKEN Genomic Sciences feedbacks could also indicate an emergence of homeostasis Center, Kanagawa, Japan; 3Sony Computer Science in mRNA. Conversely, if an expression level exceeds at least Laboratories, Inc., Tokyo, Japan one of two thresholds, the gene can be considered significantly expressed. Therefore, the two peaks could be applicable to the Objective: Intracellular dynamics of mRNA characterizes the quantitative criteria for significant change of gene expressions in cellular activity, and the behavior of individual mRNA has mainly both up and down directions. been researched as “local” gene expression for the genetic and Dedicated signaling pathway. However, intracellular “global” dynamics of DS2-2-146 Posters mRNA has rarely investigated. In this study, we analyzed a variety of temporal microarray data of human cells, and attempted Quantum-mechanical description of step-by-step protons to ascertain common properties of “global” gene expression. motion through membrane protein channels Although microarray data generally includes the uncertainty Mashkovtseva, Elena; Boronovsky, Stanislav; Nartsissov, Yaroslav because of the noise, we tried to overcome such obstacles by Institute of Cytochemistry and Molecular Pharmacology, Moscow, constructing a stochastic model. Russian Federation Result: In time-series experiments with microarrays, the highest and lowest expression values in each gene look like the random Objective: Due to van Grotthus mechanism of bonds wire, the occurrence even though several characteristic genes certainly motion of protons cannot be considered in physical terms of show the extremely high expression in the proper time. However, classical diffusion. Nevertheless, the motion of such a particle investigating when and how many genes take the maximum couples different types of the processes and could be the base of and the minimum values, we showed the fact that distributions an explanation of biological phenomena. In particular it is widely of maximum and minimum gene expression were symmetrical known that the process of proton motion through the membrane each other. Such a symmetrical aspect was confirmed in time- protein pores of ATP synthases is tightly coupled with energy series experiments of various human cells, which had publically storing into the chemical bond of ATP. However, the key point of been released as the ArrayExpress database. This symmetrical direct energy transfer still remains unclear. distribution was considerably sensitive to the normalization Results: In the present study we developed a computer model of method of microarray, so that we invented a stochastic model of step-by-step motion of protons through the limit inner cavity of a intracellular “global” mRNA dynamics in order to evaluate whether protein pore. The main feature of this model is the consideration the normalization between two arrays was well-designed. In fact, of proton motion through a channel as a succession of jumps when temporal microarray data were properly normalized each between spherical potential wells. The potential barrier can be other, this symmetrical distribution distinctly appeared. chosen as the energies need for of O-H or N-H bonds breakage Conclusion: Such symmetrical aspect in gene expression and this value is modified by any external electric field. The was observed in all of temporal microarray data, which we had probability of the jumps can be obviously calculated on the base analyzed so far. Therefore, we conclude that it seems possible of wave functions, and a mean time of the jump is estimated that this symmetrical distribution is a common feature on arbitrary using mean value of momentum operator value. Protons lost live human cells. Although the detailed intracellular mechanism their energy during interactions with the protein residues and behind this symmetry is not perfectly understood, our stochastic molecules of water in protonation/deprotonation processes and model implies that markedly up-regulated genes play major if the energy has been dropped out to the protein residue of ATP roles, and that down-regulated genes are strongly affected by synthase then it will be transmitted and stored finally into the stochasticity in the cell. energy of ATP bond. In other cases the energy will be converted into a heat. The portion of energy for a singly transaction within DS2-2-145 outlet channel of ATP synthase in absence of external electric field can be different but it would be presumably ranged between 0.2 Non-trivial peak phenomenon in distributions of maximums *10-20 and 70*10-20 J. and minimums in global gene expression analysis from Conclusion: Different abilities to combine with water or protein time series microarray data residues could partially explain a various number of protons need Sazuka, Naoya1; Asakawa, Takeshi1; Shiraishi, Tetsuya2 for a single ATP molecule synthesis in this case. 1Sony Corporation, Tokyo, Japan; 2Sony Computer Science Laboratories, Inc., Tokyo, Japan

Objective: Microarray makes it possible to measure tens of thousands gene expression values simultaneously and has been widely used as an effective technology to understand gene transcripts system quantitatively. However, a statistical and quantitative property of global gene expression dynamics is not yet well understood. The objective of this paper is to extract a statistical property of gene transcripts system from microarray

160 ICSB 2008 DS2-2-147 accuracy achieved by ZDP1 in the examples is highly satisfactory. Further, ZDP has the advantage of having a simple definition Dynamic patterns in the simulated evolution of gene and the derivation of reduced systems is straight-forward. The regulatory networks analytical expressions for the reduced system can, however, be Steiner, Till; Jin, Yaochu; Sendhoff, Bernhard prohibitively involved. In such cases, ZDP approximations of SIMs Honda Research Institute Europe GmbH, Offenbach/Main, can be calculated numerically. Germany DS2-2-149 Objective: We present a method for the in silico simulation and analysis of evolutionary multicellular development. This method Integrated approach for modelling physiological, combines simple cellular models with an abstracted DNA to biomechanical, and molecular-genetic aspects of human control cellular behavior. Simulated evolution yields genotypes cardiovascular system in health and essential hypertension adapted to the task of controlling stable growth. The in silico Semisalov, Boris1; Leonova, Tatyana1; Biberdorf, Elina2; Sharipov, nature of the approach allows for comprehensive analysis of both, Ruslan3; Yevshin, Ivan3; Trakhinin, Yuriy2; Puzanov, Mikhail1; the dynamics on the GRNs of evolved solutions, as well as the Blokhin, Alexandr2; Markel, Arkadiy4; Ivanova, Lyudmila4; evolutionary history of such dynamics (dynamic motifs). Kolpakov, Fedor1 Results: A sample evolutionary run shows that stable multicellular 1Institute of Systems Biology; Design Technological Institute of growth evolves in the framework. Analysis reveals a dynamic Digital Techniques SB RAS, Novosibirsk, Russian Federation; motif inside evolved GRNs: inter-gene negative feedback, which 2Sobolev Institute of Mathematics SB RAS, Novosibirsk, Russian emerges during evolution and spreads throughout the population Federation; 3Institute of Systems Biology; Institute of Cytology and without being explicitly selected for. Furthermore, simulated Genetics SB RAS, Novosibirsk, Russian Federation; 4Institute of mutational and knockout experiments show that the reason for Cytology and Genetics SB RAS, Novosibirsk, Russian Federation the selective advantage of this motif is the robustness of growth resulting from it. Objective. Essential hypertension (EH) and induced pathologies Conclusion: Combining computational models for the evolution of cardiovascular system (CVS) are the leading causes of mortality of GRNs with systems biology research can give new insight into in developed countries. Because EH is a complex multifactorial biological processes on an evolutionary scale, which otherwise disease, delicate approaches for treatment and prophylaxis seems not possible due to the sparseness of the fossil record. should be used. Until the present, there is no consensus of The role of dynamic patterns or motifs can be identified and opinion, general approach or catalogue, which would describe understood in computer simulations, and thus be used to this formally. The aims of the work were: 1) computational augment investigations in biology. For example, the stabilizing modelling tools improvement; 3) design of an integrated CVS Posters role of inter-gene negative feedback could now be investigated model on the base of mathematical and biological models Dedicated in vivo, and this investigation could yield refined data for the describing hemodynamics and blood pressure (BP) regulation in computational model, such that further experiments could be normal and hypertensive state; 4) discrimination of CVS model’s used to enhance the understanding of the evolution of negative key nodes critical for EH development with further verification in feedback in biology. experiments. Results. The united model of human CVS was created on the DS2-2-148 base of 1D model of arterial blood flow [1], and two different models of blood circulation system [2, 3]. BioUML (http:// Simplifying dynamics: The ZDP method www.biouml.org) was used for modelling data annotation and Härdin, Hanna1; Zagaris, Antonios2 integration. Methods of lines and orthogonal marching were used 1VU University Amsterdam, Molecular Cell Physiology, to improve mathematical apparatus for 1D model. Numerical data Amsterdam, Netherlands; 2Universiteit van Amsterdam, Korteweg for blood flow dynamics were obtained for all arteries and cross- - de Vries Instutuut, Amsterdam, Netherlands sections examined. Conclusions. The united model was tested. It allows to track Objective: Molecular processes within living cells proceed on dynamics of blood pressure in different conditions in any of 55 a wide range of timescales. In the state space of a biochemical main human arteries within a second or for days. Obtained results model, this causes the molecular concentrations to quickly move are in a good agreement with literature data. towards, and subsequently reside close to, a subspace called a Availability. Materials of the biological part of this project are slow invariant manifold (SIM). Therefore, a simplified description available in the BMOND database at http://bmond.biouml.org. of the system can be obtained by only taking the dynamics within Acknowledgements. This work was supported by integration the SIM into account. A widely used method to perform such and interdisciplinary grant 46 of SB RAS. model reduction is the quasi-steady state approximation (QSSA) References. which is highly useful due to its simplicity, although sometimes 1. D.N. Lamponi. One dimensional and multiscale models for not sufficiently accurate. Our objective was to investigate the blood flow circulation. Pour l’obtention du grade de docteur es biochemical usefulness of a recently developed timescale sciences. EP, Lausanne, 2004. reduction method based on the zero-order derivative principle 2. F.Karaaslan et al. (2005) Long-term mathematical model (ZDP). This method provides a series of approximations out of involving renal sympathetic nerve activity, arterial pressure, and which the first (ZDP0) is equivalent to QSSA, and the succeeding sodium excretion, Ann Biomed Eng., 33: 1607-1630.

(ZDP1, ZDP2, ...) can be seen as consecutive refinements of the 3. A.P.Proshin, Y.V.Solodyannikov (2006) Mathematical Modeling approximation compared to the preceding ones. The method of Blood Circulation System and Its Practical Application, has up to now only been applied to small example systems and Automation and Remote Control, 67(2): 329-341. apparently never before to a biochemical system.

Results: We show the results of applying ZDP1 to biochemical DS2-2-150 models and compare with the QSSA. For a model of an enzyme- catalyzed reversible reaction, reduced models were determined Do genome melting maps define the signature of analytically. The ZDP1 gave, as expected from theoretical results, nucleosome-depleted sites in promoter regions? a better approximation of the dynamics than the QSSA, which Hovig, Eivind1; Joshi, Himanshu N2 was particularly notable when using a parameter set for which the 1Rikshospitalet-Radiumhospitalet Medical Center, Department of QSSA gave a very poor result. For a large model describing the Tumor Biology, Oslo, Norway; 2Chalmers University of Technology, glycose phosphotransferase system in E. coli, for which achieving Gøteborg, Sweden an analytic expression for the reduced system is not feasible (for

ZDP1 as well as QSSA), we provide numerical approximations of The genomic melting map, which reflects the local SIMs based on ZDP. thermodynamic stability and propensity for single-strandedness, Conclusions: For purposes of biochemical modeling, the is an important element in understanding the processes of

ICSB 2008 161 transcriptional initiation and its transition to elongation. It is also Kholodenko [1] proposed a set of 16 basic designs that differ in known that sequence-preferences for the stable nucleosome their feedback structure. We carry out a clustering analysis to see positioning are influenced by local thermodynamic context. how the network topology affects robustness and sensitivity in Objecitive : We aimed to examine for possible coordination order to shed more light on low-level design principles to establish in the mechanisms of melting properties with other structural robust complex functionality. properties responsible for chromatin organization. Here we focus Results: We compare ODE models of 16 different topologies on the structural property of stable nucleosome positioning of a single-site phosphorylation/dephosphorylation cycle with located in putative transcriptional regulatory region (-1 to +1 inter-coupled positive and negative feedback loops proposed kbp relative to the TSS) and present a statistical comparison of in [1], each of which can exhibit relaxation oscillations. Monte melting characteristics of stable nucleosome-depleted flat melting Carlo methods are employed to explore each model’s parameter segments and that of stable nucleosome-overlapping flat melting space, and local analysis of state, period, phase and amplitude segments from a subset of human promoters. sensitivities is performed on the sampled parameter sets. The Results : Average nucleosome occupancy was higher in 0 local data is normalized and accumulated to yield semi-global to +200 bp downstream of the TSS when compared to the measures of certain robustness properties suitable for cross- occupancy in -300 to -100 bp upstream ( p< 10e-16).In addition model comparisons. Hierarchical clustering methods are then to the clustering of flat melting segments and transcription employed to visualize (dis-)similarities between different models factor binding sites in the vicinity of the transcription start site, with respect to the obtained data, using, amongst others, the we observed sufficiently lower melting temperature in the stable Jensen-Shannon divergence and Spearman’s rank correlation nucleosome-depleted flat melting segments. coefficient as distance measures. Conclusion : Based on the preliminary statistical analysis, we Conclusions: We show that hierarchical clustering based on propose that genomic melting maps may define a signature of certain robustness measures, and with the Jensen-Shannon potential nucleosome-depleted sites in promoter regions. These divergence as distance metric, groups topologically more similar nucleosome-depleted sites in association with flat melting models closer together, maintaining a high cophenetic correlation, segments with optimum melting temperature may help to identify thus supporting the robustness measures’ significance. the sites of transcriptional bubble formation. Further algorithmic Furthermore, we find that the Jensen-Shannon divergence yields developments for prediction of nucleosome positioning, in higher cophenetic correlation than Spearman’s rank correlation Dedicated combination with the experimental exploration, is necessary coefficient. Posters to define any orchestration of nucleosome positioning and [1] B. N. Kholodenko (2006) Cell-signalling dynamics in time and the optimum melting temperature of flat melting segments in space. Nat. Rev. Molec. Cell Biol. 7: 165-176. promoter regions. DS2-2-153 DS2-2-151 Oscillations in the expression of a self-repressed gene Flux changes in myocardial energy metabolism under induced by a slow transcriptional dynamics external perturbation Lefranc, Marc1; Morant, Pierre-Emmanuel1; Thommen, Quentin1; Hasan, Md. Nabiul; Oh, Eulsik; Yoo, Young Sook; Lemaire, François2; Parent, Benjamin3; Vandermoëre, Constant1 The Korea Institute of Science and Technology, Life Sciences 1University Lille 1/CNRS, PhLAM, Villeneuve d’Ascq, France; Research Division, Seoul, Republic of Korea 2University Lille 1/CNRS, LIFL, Villeneuve d’Ascq, France; 3University Lille 1/CNRS, UGSF, Villeneuve d’Ascq, France Recently, considerable attentions have been devoted to the chronically failing heart. Also, it has been suggested that the Objective: show in a simple example that the existence of a chronically failing heart is related to metabolically abnormal transcriptional dynamics, starting with the fact that transcription states of the heart. Generally, the healthy heart gets its energy rate does not react instantaneously to a change in a regulator to beat from ATP synthesis by oxidative phosphorylation of fatty protein, must be taken into account in dynamic models of gene acid oxidation. Thus, if the balance of the energy metabolism is circuits. broken, the healthy heart has to beat in an energy-starved state. Results: Recent experiments have shown the existence of In this research, we studied the flux changes of metabolites in an autonomous transcription rate dynamics at time scales myocardial energy metabolism using a computational model of comparable to other important processes including degradation. mouse cardiac myocyte when external perturbations, a deletion of We have investigated the dynamical consequences of a finite- reaction in the metabolic pathway, are given. The deletion is given time gene response in the behavior of the simplest genetic circuit, in two different ways. One is an untargeted deletion. The other is consisting of a gene whose expression is repressed by its own targeted deletion. After the changes, we examined the changes protein. This system is known to oscillate when a time delay is of pathways and metabolites to unveil the biological properties of inserted in the feedback loop or when degradation mechanisms energy metabolism in mouse cardiac myocyte. We founded that are strongly nonlinear but it has generally been assumed that the cardiac system did not make a roundabout way under the transcription rate reacts instantenously to protein concentration. targeted deletion so that the pathway is considerably broken. It We have found that the appearance of sustained oscillations suggested that the energy metabolism in mouse cardiac myocyte requires much less nonlinear degradation mechanisms than in was not robust in the targeted attacks. the infinitely fast regulation case. We have obtained analytical expressions for the oscillation threshold which allow us to identify DS2-2-152 the key oscillation parameters and to evidence a resonance phenomenon where the one-gene circuit is most strongly Topology and robustness: A clustering analysis of destabilized when gene response time crosses a critical value Kholodenko’s minimal oscillator models whose expression can also be given analytically. Franjcic, Zlatko; Dubhashi, Devdatt Conclusion: a non-trivial transcriptional dynamics modifies the Chalmers University of Technology, Computer Science and dynamical behavior of a simple gene circuit and thus must be Engineering, Gothenburg, Sweden taken into account in theoretical models.

Objective: Biochemical networks in living systems have to perform a variety of functions, some of which are highly complex. Ordinary differential equation (ODE) models are commonly used to describe and analyze the temporal dynamics of such network functions. A particularly interesting class of models is the one of oscillatory systems. This kind of behavior has traditionally been attributed to cascade networks. Yet even models of single-cycle networks can, under certain conditions, exhibit oscillations.

162 ICSB 2008 Dedicated session 2-3: Diagnostic Objective: The aim of this study was to use a proteomic markers and complex diseases approach for identification of novel biomarkers that could be used to discriminate between high and low GH treatment responders in short prepubertal non GH-deficient children. DS2-3-09 Sera from 32 children, 13 with high GH treatment response (0.9- 2.0 Δheight SDS during 1st yr) and 19 children with low treatment Surgical supersystems: Metabonomic profiling of the gut response (0.2-0.5 Δheight SDS) were analysed before and after microbiome during surgically induced intestinal ischaemia 1st year on GH treatment, using the protein profiling technique / reperfusion injury SELDI-TOF. Analysis was performed using weak cation exchange Kinross, James1; Barton, Richard2; Alkhamesi, Nawar3; Hunte, (CM10) arrays. Cross-validated stepwise regression models were Kerrilyn4; Tuohy, Kieran4; Silk, David5; Holmes, Elaine6; Darzi, Ara5; used to investigate if a combination of biomarkers could be used Nicholson, Jeremy2 for discrimination between high and low responders. 1Imperial College London, Department of Biosurgery and Surgical Result: In this study we identified two peaks of 13.8 and 17.1 Technology, London, United Kingdom; 2Imperial College London, kDa from which changes in peak intensities from start to 1 year Biomolecular Medicine, London, United Kingdom; 3Imperial of treatment together could be used to discriminate between high College London, Biosurgery and Surgical Technology, London, and low GH treatment responders with a correct classification United Kingdom; 4Reading University, Food Biosciences Unit, rate of 0.82 (R2=0.47). The 13.8 kDa peak was related with Reading, United Kingdom; 5Imperial College London, Biosurgery auxological data and the 17.1 peak was related to the GH/IGF-I and Surgical Technology, London, United Kingdom; 6Imperial axis. By the consistency of the peak pattern in spectra, depletion College London, Biomolecular Medicine, London, United experiments and MS protein identification, these peaks have Kingdom been identified to represent the nutrition factors transthyretin and apolipoprotein A-II, respectively. Objectives: Surgically induced intestinal Ischaemia / Conclusion: From this study we conclude that a proteomic Reperfusion (I/R) injury drives postoperative sepsis and Multi approach can be used to discriminate between high and low Organ Dysfunction Syndrome (MODS). The gut microbiota are GH treatment responders in short prepubertal non GH-deficient implicated as an important co-factor in this mechanism. A suite children. Furthermore, new knowledge about GH-dependent of metabonomic and genomic technologies were therefore proteins can be used to better understand the mechanisms employed to determine how surgically induced I/R injury disrupts behind GH responsiveness. the gut microbiome and to identify novel in vivo biomarkers that describe its influence on post operative outcome. A standardised DS2-3-11 non recovery superior mesenteric artery I/R model was developed Posters using 27 male Wistar rats: Sham (n = 6), 30 mins (n = 7), 60 The “prion strain phenomenon” investigated through a Dedicated mins (n = 8) and 90 mins (n = 6) of ischaemia. Animals were kinetic replication model exposed to 90 mins of reperfusion. Urine and plasma were Altafini, Claudio; Zampieri, Mattia; Legname, Giuseppe analysed by a Bruker 600Mhz 1H Nuclear Magnetic Resonance SISSA Int. Sch. for Advanced Studies, Trieste, Italy (NMR) spectroscope. Data reduction was performed for use in SIMCA software and for direct import into Matlab where in- Objective: Prions are infectious agents solely composed of house multivariate methods were applied. Stool samples were proteins whose replication does not rely on the presence of analysed by polymerase chain reaction (PCR), and Denaturing Gel nucleic acids and whose proliferation induces neurodegenerative Electrophoresis (DGGE) of microbial 16S rDNA. diseases. Their ability to misfold into a range of different Results: Principal Components Analysis (PCA) of urine spectra aggregated forms, enciphering a multitude of phenotypic significantly classed animals according to ischaemic severity (R2 states, is known as “prion strain phenomenon”. Recent studies = 0.769, Q2=0.599) and I/R status (R2 0.707=, Q2 = 0.446). demonstrate that the disease incubation time is directly Pseudospectral loading plots from Orthogonal—Partial Least proportional to the relative resistance of the prion isolates against Squares-Discriminant Analysis (O-PLS-DA) showed microbial denaturation. More stable prions lead to large vacuolation co-metabolites such as hippurate were significant in making I/R targeting specific brain regions, while infection with unstable class description possible (R2 = 0.917, Q2 = 0.646) and were prion strains presents less intense and more widely distributed negatively predictive for the reperfusion state. Bifidobacteria vacuolation in the brain. specific and archaea-specific PCRs of stool samples were Results: In this study we investigate the relationship between negative. PLS-DA analysis of the DGGE bands identified clear and the stability against denaturation and other empirical parameters consistent differences in the intestinal ecosystem between sham (derived from in vivo experiments on mammals) describing the animals and those exposed to 30, 60 and 90 (R2 = 0.699, Q2 = kinetics of prion replication. In order to do this we use a well 0.79) minutes of ischaemia. established mathematical model describing the fibril length Conclusion: A metabonomic strategy can objectively quantify evolution. The main observation we make is that the stability to altered microbial metabolism during surgically induced I/R injury, denaturation is directly proportional to the reproductive ratio and and this occurs with significant fluctuations in the local microbial inversely correlated to the rate of growth. In terms of the kinetic ecosystem. model, the key parameter describing prion strain dependent replication kinetics seems to be the fibril breakage rate. DS2-3-10 Conclusions: The analysis points out that stable prion strains are characterizable by means of a “stronger” aggregated structure, A proteomic approach identified growth hormone which also implies a longer fibrils mean length. On the other hand, dependent nutrition markers in children with idiopathic unstable prion strains are more prone to breaking, resulting in short stature shorter fibrils. Hellgren, Gunnel1; Andersson, Björn1; Nierop, Andreas2; Dahlgren, This model-based analysis is consistent with what has been Jovanna1; Hochberg, Ze’ev3; Albertsson-Wikland, Kerstin1 experimentally observed in a simpler unicellular organism like 1Institute of Clinical Sciences, Department of Pediatrics, yeast. Gothenburg, Sweden; 2Muvara bv, Leiderdorp, Netherlands; The characterization of the prion strains in terms of polymer mean 3Meyer Children’s Hospital, Rambam Medical Center, Haifa, Israel size provides also a possible explanation to the fact that stability is also related with lesion profiles, by means of vacuolation areas: The regulation of longitudinal growth is a complex process the increased size associated to stable prions can decrease governed by the interplay between numerous factors and their ability to diffuse, and can circumscribe them in small brain pathways, which of many are still unknown. In this study we regions. On the contrary, oligomers can spread around the brain have used protein profiling (SELDI-TOF) in order to improve the more easily causing a more homogeneous damage. understanding of underlying mechanisms involved in growth hormone (GH) responsiveness.

ICSB 2008 163 DS2-3-13 hundreds of lipid molecular species. Such data then leads to several analyses, such as multivariate exploratory analyses and The cellular level of cardiovascular diseases: Erythrocyte correlation network analyses, as well as combined analyses of aspect lipid and gene expression profile data Brazhe, Nadezda1; Luneva, Oksana2; Bryzgalova, Nadezda2; Conclusions:The methodology facilitates identification and Rodnenkov, Oleg3; Parshina, Evgeniya2; Abdali, Salim4; interpretation of high-throughput lipidomics data. In the context of Sosnovtseva, Olga1; Mosekilde, Erik1; Maksimov, Georgy2; multi-tissue lipidomic profiling in POKO mouse, we will show most Chazov, Evgeniy3 discriminating variables as well as linking the information to lipid 1The Technical University of Denmark, Department of Physics, pathways using previously demonstrated informatics strategies BioSim group, Kgs. Lyngby, Denmark; 2Moscow State University, Biophysics Department, Biological Faculty, Moscow, Russian DS2-3-16 Federation; 3Russian Cardiological Research Complex, A.L. Myasnikov Institute of Clinical Cardiology, Moscow, Russian Local peak finding for ChIP-seq data corrects inherent Federation; 4The Technical University of Denmark, Department of bias in sequence alignment and facilitates the discovery of Physics, Kgs. Lyngby, Denmark epigenetic markers Choi, Jung Kyoon; Lyu, Jaemyun; Cho, Hwan Sung; Bae, Jae- Objective: The common view on the origin of cardiovascular Bum; Kim, Young-Joon diseases (CVD) is the pathology arises on the tissue and organ Yonsei University, Seoul, Republic of Korea level, for example, in the heart, vessels or capillaries. However, there is also a cellular aspect related to erythrocytes. Distortion Objective: With the recent advent of Solexa sequencing of erythrocyte function can cause a reduced oxygen delivery technology, we see rapid growth of epigenomics data. Unique to tissues and can produce additional hypoxia. The aim of our features of the ChIP-seq data call for new analysis methods, project is to study the effect of various cardiovascular diseases which should be different from those designed for ChIP-chip and drugs on erythrocyte properties: plasma membrane fluidity, data. In particular, we observed strong inherent bias in sequence activity of Na+/H+-exchanger and Ca2+-ATPase and the ability of alignment for repeat-enriched genomic regions. This led us hemoglobin to bind and release oxygen. to develop a local peak-finding algorithm which identifies Dedicated Results: As experimental setups we use Raman and EPR- consecutive sequence reads that are statistically significant in Posters spectroscopies and ion-selective electrodes. Obtained data show comparison to background region. that the plasma membrane fluidity of erythrocytes, the activity of Results: Our method was first tested for public data on histone membrane proteins, and the oxygen-binding ability of hemoglobin modifications in mouse embryonic stem cells. We then analyzed are changed under heart failure, ischemia, essential hypertension our DNA methylation data from mouse cancer tissues. Since and administration of drugs, based on the nitric oxide donors. our algorithm produces a normalized score for each locus from Natural hypoxia from birth (mountain hypoxia) also affects each experiment, we successfully detected hyper- and hypo- erythrocytes properties, but found changes are seemed to be methylation sites in tumor cells, which presented distinctive adaptive to hypoxic conditions. features of cancer-induced CpG methylation. In contrast to Conclusions:We propose that the observed changes in consistent hypomethylation signals observed over long repeat properties of erythrocytes of CVD patients can aggravate hypoxia regions, hypermethylation sites tend to form local peaks that are and should be considered in treatment of patients and during the generally found in genic regions. Interestingly, intragenic regions, testing of newly developed drugs. rather than promoters, were the major target of CpG methylation, implying a role for the regulation of transcription elongation. We DS2-3-15 report some of the genes that have tumor-specific peaks in their promoters or gene bodies. Lipid informatics approaches for phenotype Conclusions: Our method will serve to detect normalized, bias- characterization using lipid pathways free local peaks from ChIP-seq data for chromatin modification Yetukuri, Laxman1; Katajamaa, Mikko1; Medina-Gomez, Gema2; and CpG methylation. It will be especially useful for scanning Seppänen-Laakso, Tuulikki1; Vidal Puig, Antonio2; Orešič, Matej1 differential epigenetic patterns for marker discovery, which has 1VTT Technical Research Centre of Finland, Espoo, Finland; been a major application of ChIP-chip analysis but not been 2University of Cambridge Department of Clinical Biochemistry, established for ChIP-seq data. Cambridge, United Kingdom DS2-3-17 Objective:Phenotype of a complex pathobiological state is characterised by integrated study of differential fold changes 17p11.2 duplication/deletion analysis in CMT1 patients in at genomic and lipidomic (or metabolomic in general) levels Ukraine in response to genetic or environmental perturbation. Current Hryshchenko, Nataliya; Livshits, Ludmila lipidomic studies, however, needs strategies for the elucidation Institute of Molecular Biology and Genetics NASU, Human of important pathobiological phenomena from the integration of Genomics, Kiev, Ukraine the large amounts available data. Here we present computational and informatics approaches to study lipid molecular profiles Objective: Detection and analysis of 17p11.2 chromosome in the context of known metabolic pathways and established region rearrangements in CMT1 patients from Ukraine. pathophysiological responses, utilizing information obtained from Results: Two intercomplementary methods of 17p11.2 modern analytical technologies duplication/deletion identification have been elaborated: STR Results:In order to facilitate identification of lipids, we compute allelic varians analysis and direct PMP22 gene dosage measuring the scaffold of theoretically possible lipids based on known lipid by means of quantitative Real-Time PCR. It has been carried building blocks such as polar head groups and fatty acids and out detection and analysis of 17p11.2 chromosome region utilize recently developed nomenclature of lipids.Each compound rearrangements in CMT1 patients from Ukraine. CMT1A- entry is linked to the available information on lipid pathways and duplication have been identified in 19 families with clinical CMT1 contains the information that can be utilized for its automated symptoms including 2 families with de novo mutation. In 2 CMT- identification from high-throughput UPLC/MS-based lipidomics patients with specific clinical symptoms (father and son from one experiments. Our global profiling lipidomic platform involves family) HNPP-deletion have been identified. screening of major lipids, including acylglycerols, phospholipids, Conclusions: It has been registered the high level of de novo sphingolipids, and cholesterol esters. Several data processing cases with 17p11.2-duplication. It has been sown the 17p11.2 steps such as peak detection, alignment, and normalization chromosome region duplication/deletion association with CMT1A are performed using MZmine software. The resulting peaks are and HNPP clinical phenotypes which may be used in differential identified utilizing a comprehensive spectral library of lipids, which diagnosis of this type of CMT polyneuropathy. afford automatic identification and profile comparison of several

164 ICSB 2008 DS2-3-18 DS2-3-20

Multi-agent targeted therapy of cancer: a reverse The study of SNP’s in FSH receptor and INHa1 genes in engineering approach women with diminished ovarian reserve from Ukraine Nelander, Sven; Nelander, Sven Livshyts, Ganna; Podlesnaja, Svetlana; Kravchenko, Sergey University of Gothenburg and Memorial Sloan Kettering Cancer Institute of Molecular Biology and Genetics (National Academy Center, Gothenburg, Sweden of Sciences of Ukraine), Department of Human Genomics, Kiyv, Ukraine We present a novel method for deriving network models from molecular profiles of perturbed cellular systems. The network Objectives: Premature ovarian failure (POF) is a secondary models aim to predict quantitative outcomes of combinatorial gonadotrophic amenorrhoea affecting 1-3% of females. FSH perturbations, such as drug pair treatments or multiple genetic (follicle stimulation hormone) and its receptor (FSHR) play a major alterations. Mathematically, we represent the system by a set role in the development of follicles and regulation of steriogenesis of nodes, representing molecular concentrations or cellular in the ovary. Mutations in the FSHR might theoretically lead to processes, a perturbation vector and an interaction matrix. After an impaired signal transduction and thereby to a diminished perturbation, the system evolves in time according to differential ovarian reserve. Genes encoding the three inhibin subunits equations with built-in non-linearity, similar to Hopfield networks, can be proposed as candidates for POF due to its role in the capable of representing epistasis and saturation effects. For a negative feedback control of FSH. We investigate Asn680Ser and particular set of experiments, we derive the interaction matrix Thr307Ala transitions in FSHR gene and Ala257Thr transition in by minimizing a composite error function, aiming at accuracy of INHá1 gene as markers for diminished ovarian reserve in women: prediction and simplicity of network structure. To evaluate the with clinical POF diagnosis and women with poor response (less predictive potential of the method we performed twenty-one than 4 oocytes after standard protocol of FSH stimulation) -”poor drug pair treatment experiments in a human breast cancer cell responders”. line (MCF7) with observation of phospho-proteins and cell cycle Results: The frequency of Ala307-Ser680 (AS) allele of FSHR markers. The best derived network model rediscovered known gene was significantly higher both in POF group and in “poor interactions and contained interesting predictions. Possible responders” group comparing to control group. The carriers applications include the discovery of regulatory interactions, the of INHa1 gene Ala257Thr transition predominated in the “poor design of targeted combination therapies, and the engineering of responders” group. Quantity of oocytes after stimulation of molecular biological networks. ovulation in women with INHa1 gene Ala257Thr transition was significantly decreased in comparison to patients without such

DS2-3-19 mutation. Our data shows the prevalence of FSHR gene AS allele Posters in both patients groups: group of POF patients (45,7%) and “poor Dedicated Gene expression analysis of various cancers using a new responders” group (52,8%), comparing to control group (35,1%). method combined with boosting and projective adaptive Conclusions: Our data about FSHR and INHa1 genotype resonance theory association with ovarian reserve and response to FSH represent Takahashi, Hiro1; Honda, Hiroyuki2 that the best stimulation protocols can be based on the 1Chubu University, College of Bioscience and Biotechnology, individual’s genetic profile in order to reduce side-effects and Kasugai, Japan; 2Nagoya University, Department of costs and improve the delineation stimulation protocols. Biotechnology, School of Engineering, Nagoya, Japan DS2-3-21 An optimal and individualized treatment protocol based on accurate diagnosis is urgently required for the adequate treatment Training-induced effects on plasma metabolomics in of patients. For this purpose, it is important to develop that a patients with chronic obstructive pulmonary disease sophisticated algorithm that can manage large amount of data, Alcarraz-Vizán, Gema1; Rodríguez, Diego2; Selivanov, Vitaly1; such as gene expression data from DNA microarray, for optimal Reed, Michelle3; Gómez, Federico2; Günther, Ulrich3; Roca, and individualized diagnosis. Especially, marker gene selection is Josep2; Cascante, Marta1 essential in the analysis of gene expression data. 1University of Barcelona, Biochemistry and Molecular Biology, In the present study, we developed the combination method Barcelona, Spain; 2Hospital Clínic, IDIBAPS, Barcelona, Spain; of projective adaptive resonance theory and boosted fuzzy 3CR UK Institute for Cancer Studies, University of Birmingham, classifier with SWEEP operator method (PART-BFCS method) Birmingham, United Kingdom for model construction and marker selection. We also applied this method to microarray data, such as brain tumor and acute Objective: Patients with chronic obstructive pulmonary disease leukemia. The method enabled the selection of 14 important (COPD) show abnormal adaptations of skeletal muscle redox genes related to the prognosis of the tumor for brain tumor. status after training. The phenomenon is clearly more evident In addition, we proposed improved reliability index for cancer in COPD patients with muscle wasting. The analysis of small- diagnostic prediction of blinded subjects. Based on the index, molecule metabolite profiles (metabolomics) may shed further light the discriminated group with over 90% prediction accuracy was on the adaptations of COPD patients to training. We aimed at separated from the others. Therefore, this method was applied to analyzing plasma metabolomic changes after 8-week endurance other gene expression data obtained from various cancers such training. Methods: We studied 13 patients with stable COPD and as, esophageal cancer, soft tissue sarcoma, and lymphoma. As normal fat free mass (COPD_FFMN 68±4 yrs, FEV1 49±8% pred), a result of these analyses, we could select many marker gene 6 COPD patients with low FFM (COPD_FFML 69±11 yrs, FEV1 candidates, such as CDK6 for esophageal cancer, MIF for soft 41±16%) and 12 health sedentary controls (H 65±9 yrs, FEV1 tissue sarcoma, and FES and GADD45A for lymphoma. 107±14%). Pre- and post-training plasma samples at rest and

PART-BFCS with improved RIBFCS method does not only show after constant-work rate exercise at 80% of pre-training VO2peak high performance, but also has the feature of reliable prediction were analyzed by 1H-NMR spectroscopy. Principal component further. This result suggests that PART-BFCS with improved RIBFCS analysis (PCA) was done. Metabolic effects of training were method has the potential to function as a new method of class studied using a kinetic model of central metabolism. prediction for diagnosis of patients. Results: Training-induced enhancement of both VO2peak and 6MWD was seen in all groups (p<0.01). At rest, PCA revealed

different metabolomic patterns among COPD_FFMN, COPD

FFML and H. In all cases, acute exercise (pre- and post-training) generated clear metabolomic changes. While PCA showed

similar training-induced metabolic profile between COPD_FFMN

and H, COPD_FFML did not present significant metabolomic improvement. Model analysis revealed that training induces more

ICSB 2008 165 effective oxygen metabolism in healthy individuals than in severe Utilizing a collection of predicted microRNA (miRNA) targets, we COPD patients. identified some motifs that were disproportionately represented Conclusions: Metabolomics differentiates the 3 groups at rest in these gene sets, suggesting hypothetic regulatory networks

and indicates that training improves COPD_FFMN but not COPD_ for the observed gene expression patterns. Such analysis was

FFML response to exercise. Integration of clinical information with performed by applying a hypergeometric test, and adjusting other “omics” data (functional genomics and proteomics) may p-values with Benjamini-Hochberg correction. help to understand the underlying mechanisms of muscle wasting Differentially expressed genes were then projected on the in COPD and perform early identification of those COPD patients modules obtained from clustering of the mouse Protein Interaction with poor prognosis because of systemic effects. (Supported by Network (PIN). Such modules may be thought of as groups grants from BioBridge (FP6-2005- 037909) FIS (2005-061510) of tightly interacting proteins, engaged in the similar biological and ERS-SEPAR fellowship grant #191). process. Several PIN modules were enriched in ‘presence of differentially expressed genes’, indicating that functions DS2-3-22 represented by these modules are involved in the formation of cancer. Measuring similarities between genetic alteration patterns We also searched for the processes whose regulation becomes in cancer cells disrupted in cancer by analyzing clusters of genes that are Soneson, Charlotte1; Lilljebjörn, Henrik2; Fioretos, Thoas2; Fontes, correlated in healthy tissue but loose their correlation in the Magnus1 cancer states. 1Centre for Mathematical Sciences, Mathematics LTH, Lund, With these approaches we identified regulatory miRNAs, genes, Sweden; 2Lund University Hospital, Department of Clinical protein modules and processes associated with various cancer Genetics, Lund, Sweden development processes, including cell adhesion, cell migration, blood vessel morphogenesis, morphogenesis of the epithelium Objective:Overwhelming evidence supports the hypothesis that and apoptosis. cancer arises by a stepwise accumulation of genetic alterations in Conclusions: We identified regulatory molecules, genes and a cell. The field of cancer genetics is currently being transformed processes that are associated with the development of cancer by genome-scale technologies capable of detecting various types and demonstrated the applicability of the novel tools, like Dedicated of mutations in cancer cells. The objective of this study was to differential co-expression analysis and miRNA target prediction. Posters develop a way to measure the similarity between the presence of different such aberrations, thereby revealing any interdependence DS2-3-24 between aberrations and possibly also disclosing the functional consequence of the aberrations. Metabolic effects of altered oxygen signaling in renal Results:We propose and evaluate a way of measuring the carcinoma cells uncovered by targeted metabolomics dissimilarity between the occurrences of a pair of genetic Czernik, Dominika; Zamboni, Nicola

aberrations (xi and xj, i≠j). We let Xk denote the set of cases where ETH Zurich, Institute of Molecular Systems Biology, Zurich,

the aberration xk is present, and |Xk| the cardinality of Xk. Then we Switzerland put

D(xi,xj)=exp{sgn(α-|Xi∩Xj|/min(|Xi|,|Xj|))(min(|Xi|,|Xj|))/A-2|Xi∩Xj|/ Objective: Integration of different omics is one of the major

(|Xi|+|Xj|)} challenges in systems biology. Here we try to bridge the where A is a normalization constant and α is a tuning parameter. transcriptional and metabolome responses elicited by oxygen

We let D(xi,xi)=0. signaling in cancer cells. In normal cells, the key regulator in A good dissimilarity measure should be able to reproduce oxygen sensing is the transcriptional regulator hypoxia inducible biologically relevant relationships between genetic aberrations. factor (HIF) that promotes mechanisms compensating for oxygen Two aberrations which occur frequently in cancer cells, but rarely depletion (hypoxia), including enhanced expression of high occur in the same cell, should be assigned a high dissimilarity capacity glucose transporters and all glycolytic enzymes leading score. Conversely, two aberrations which often occur in the to lactic acid. As a result, metabolism switches from respiratory same cell should receive a small score. We apply the dissimilarity to fermentative metabolism. This mechanism is altered in most measure defined above to two datasets describing occurrences cancer cells, which exhibit anaerobic catabolism of glucose to of DNA copy number aberrations, as determined by microarray lactate even in the presence of oxygen (Warburg effect). Recent analysis, in samples from patients with childhood leukemia. studies demonstrate the causal link between this metabolic We evaluate the dissimilarity measure according to how well it alteration and cancerogenesis. A frequent cause is mutation reproduces biologically relevant relationships as discussed above, or silencing of the von Hippel-Lindau tumor suppressor protein and compare the performance to that of common distance (pVHL), which mediates the oxygen-dependent degradation measures, and dissimilarity measures based on one-sided of HIF. Inheritance of a mutant allele of the VHL provokes a p-values from a Fisher exact test. constitutive accumulation of HIF and predisposes affected Conclusions:The dissimilarity measure described above individuals to develop tumors of the kidney, pancreas, adrenal succeeds well in finding biologically relevant relationships, gland, retina and central nervous system. This makes VHL a compared to other measures. Depending on the choice of α, powerful model to study the metabolic switch in cancer cells and different similarity patterns will be revealed. identify therapeutic targets. Results: To elucidate metabolic response to perturbations of DS2-3-23 pVHL-HIF regulation, we measured the concentrations of all intermediates in central carbon metabolism comparing hypoxic Network oncology: Analysis of cancer gene expression and normoxic growth in renal carcinoma cells with either a in the context of protein interaction networks and miRNA mutated or reconstituted pVHL. Concentrations were measured target motifs with a targeted liquid chromatography-mass spectrometry Mentzen, Wieslawa; Floris, Matteo; de la Fuente, Alberto platform that covers all intermediates and cofactors of primary CRS4 Bioinformatics Laboratory, Pula, Italy metabolism. Despite the fact that hypoxia and pVHL mutation induce equivalent stabilization of HIF and thus identical Objective: Integrative systems biology approaches were applied transcriptional responses, striking and unexpected differences for the analysis of gene expression data from mouse model were detected in the metabolome. of mammary gland tumor, in the context of Protein Interaction Conclusions: The complex patterns observed between the four Network and genomic data, to identify molecular mechanisms conditions point to additional regulatory mechanism of post- involved in the progression from healthy to tumor tissues. transcriptional or allosteric nature, which will be investigated in Results: Expression data from healthy, hyperplastic, and tumor follow-up experiments. mammary tissues were used to identify sets of differentially expressed genes that mark differences between these states.

166 ICSB 2008 Dedicated session 2-4: Microbial systems catalyze those processes themselves. In addition, we developed ecological regulation analysis to quantify how biogeochemical DS2-4-09 fluxes are actually regulated by the microorganisms performing the process; to which degree are changes in fluxes due to Metabolic engineering of recombinant Escherichia coli changes in population size and changes in cellular activity. Its cells producing the antioxidant vitamin E application revealed that in general flux was regulated by cellular Lemuth, Karin1; Vallon, Tobias1; Vielhauer, Oliver1; Albermann, activity, i.e. by changes in the size and properties of the enzyme Christoph2; Ghanegaonkar, Shashank2; Sprenger, Georg2; Reuss, pool and in concentrations of substrates and metabolites. Thus, Matthias1 it is often not sufficient to count the numbers of cells performing a 1Universität Stuttgart, Institute of Biochemical Engineering, particular biogeochemical process in order to estimate its flux. Stuttgart, Germany; 2Universität Stuttgart, Institute of Conclusions: These ecological systems biology approaches, Microbiology, Stuttgart, Germany combined with molecular ecology and biogeochemistry approaches, should enhance understanding on ecosystem Objective: Vitamin E belongs to the group of lipid soluble functioning in general. vitamins and comprises the group of eight tocochromanoles (alpha - delta tocopherol and alpha - delta tocotrienol, DS2-4-11 respectively). Alpha-tocopherol shows the highest in vivo activity, however, tocotrienols are receiving increased attention Mapping high quality genome-scale metabolic network of since interesting biological activities in health and disease the mycobacterium tuberculosis with human: Applied to (neuroprotective, antioxidant, anti-cancer, and cholesterol drug target identification lowering properties) have been reported. A weak point in the Kalapanulak, Saowalak; Ma, Hongwu; Goryanin, Igor recombinant production of tocochromanoles is the availability of University of Edinburgh, School of Informatics, Edinburgh, United precursors in the cell, especially isoprenoid diphosphates. Much Kingdom empiric work has been done to enhance different enzymes in the isoprenoid pathway to improve the yield of several isoprenoids Mycobacterium tuberculosis (Mtb), a pathogenic bacterium, is like carotenoids. In order to achieve higher isoprenoid yields the causative agent in the vast majority of human tuberculosis in heterologous E. coli strains, it is necessary to quantify this (TB) cases. Nearly one-third of the world’s population has been pathway intermediates. affected by TB and annually two million deaths result from the Results: Here we show the formation of the vitamin E compound disease. Because of the cost of medication for a long period of delta-tocotrienol in recombinant E. coli cells. Furthermore, treatment with multiple drugs and the increase of a multi-drug we show the detection and quantification of the pathway resistant strain of Mtb, faster-acting drugs and more effective Posters intermediates farnesyl diphosphate (FPP) and geranylgeranyl vaccines are urgently demanded. Several metabolic pathways of Dedicated diphosphate (GGPP) in wild type (9.8 nmol*g cdw-1 FPP and 3.2 Mtb are attractive for identifying novel drug targets against TB. nmol*g cdw-1 GGPP) and recombinant strains (310 nmol*g cdw- Therefore, in this work a high quality genome-scale metabolic 1 GGPP). network of Mtb was reconstructed and compared to the human Conclusions: We have established a simple and sensitive metabolic network, published by Ma et al. in 2007, in order to nonradioactive method for the detection of isoprenoid propose new drug targets. All 666 EC numbers of Mtb network diphosphates (FPP and GGPP) in E. coli cells, which led to a were mapped to 788 EC numbers of the human metabolic recovery of > 90 % for the diphosphates used. The quantification network and compared in terms of protein functional sites through of these metabolites might be regarded as a key step in the InterPro accession numbers. There are 345 unique EC numbers metabolic engineering of isoprenoid compounds in E. coli in Mtb but only 98 of them have different protein functional sites because this enables precursors of the host strain (FPP) to be from all human proteins. These 98 EC numbers corresponding to differentiated from the first recombinant step of biosynthesis for 105 genes are interesting as drug targets. When combining the manifold isoprenoid compounds. result with a list of 614 essential genes required for optimal growth of Mtb from transposon site hybridization (TraSH) technique of DS2-4-10 Sassetti et al., only 37 from 105 genes are essential genes. In conclusion, these 37 genes are proposed to be the drug targets Control and regulation of flux through networks of which function in several pathways such as phenylalanine, interacting micro-organisms tyrosine and tryptophan biosynthesis, arabinogalactan Roling, Wilfred; Tobor-Kaplon, Maria; Westerhoff, Hans biosynthesis. Moreover, protein similarity analysis between their VU University Amsterdam, Molecular Cell Physiology, Amsterdam, gene products and all human proteins was done. The proteins Netherlands encoded by 37 genes have 22-44% identity to human proteins. We also found that 11 of them are current validated drug targets Objective: Current microbiological research on biodegradation (Mdluli and Spigelman 2006). generally focuses on the microorganisms that are directly involved in the degradation of pollutants. However, these DS2-4-12 degrading microorganisms interact with other functional groups (e.g. predators) that indirectly may have a major influence on Computational dissection of adenovirus cell surface motion degradation rates. Identification of the functional groups that reveals receptor mediated virus drifting on filopodia predominantly determine fluxes through, and concentrations in, Burckhardt, Christoph; Schönenberger, Philipp; Greber, Urs ecological networks should benefit the understanding on and University of Zürich, Institute of Zoology, Zürich, Switzerland manipulation of material fluxes. We have started addressing this by a combination of mathematical modelling and experimentation. The earliest steps of virus host cell interactions are incompletely Results: We developed ecological control analysis (ECA) to understood. In the case of human adenoviruses they involve viral quantify the control of each functional group in an ecological attachment to the coxsackie virus B adenovirus receptor CAR, network on its process rates and concentrations of intermediates. and alpha v integrins followed by endocytic uptake and infection. A theoretical study on anaerobic syntrophic degradation of CAR is a type one transmembrane protein of cell-cell junctions organic matter revealed that in contrast to current views flux and the basolateral plasma membrane domain. How the virus control can be distributed over several groups. Control over finds the appropriate domain on the plasma membrane to engage intermediate concentrations is always shared. Because of in endocytosis is unknown. Here we are using live fluorescence networking effects, the concentration of an intermediate can also imaging at high temporal resolution together with single be controlled by functional groups not producing or consuming particle tracking algorithms to systematically map the motion of it. ECA can give rise to counterintuitive results, for example adenoviruses on filopodia of human epithelial cells. Surprisingly, halorespiring microorganisms may not control the rate of the individual trajectories are of considerable heterogeneity. To perchloroethylene and trichloroethylene degradation although they unravel the information contained within these trajectories, we

ICSB 2008 167 developed a novel trajectory segmentation approach based on the subgroups in the project. The models were then integrated supervised support vector classifications. This approach revealed numerically. The integral model served as the basis for model three distinct patterns, diffusion, drifts and confined motions. driven data handling. Upon attachment to filopodia, viruses were mostly randomly The integrated systems biology model was able to explain quite diffusive, before they processively drifted towards the cell body. paradoxical effects of an agent inducing DNA damage (and repair) Adenovirus drifts and infection required filamentous actin and in yeast. Control was distributed as expected, except that the myosin2. The absence of CAR reduced Ad2 drifts whereas alpha DNA damage process had a much larger control than anticipated. v integrin depletion increased the drifts, suggesting a competition The domino approach to systems biology was demonstrated between CAR and integrins in viral surface motions. To identify in the integral model. The model will now serve as a tool for the minimal drift mediating portion of CAR we screened a number experimental design. of CAR deletion mutants, and identified a drifting domain of CAR. Conclusions: Proof of principle was given of this domino Together, our data show how a combination of live fluorescence methodology to build large models from partial models. imaging and computational analyses can reveal new features of Minor free energy draining reactions can have major control a plasma membrane receptor, which are important for infectious Multi-laboratory Systems Biology can be integrated very tightly virus-host interactions. Systems Biology data handling may need to be different from the data handling known from Bioinformatics DS2-4-13 DS2-4-15 EcoliHub - an information resource for experimentalists and modelers Quantitative phenomics and metabolic control analysis: Wanner, Barry1; Aref, Walid2; Datsenko, Kirill A.1; Ess, Sara1; Bridging the gap with yeast Gribskov, Michael R.1; Kihara, Daisuke2; Kim, Sangtae3; Mori, Gkargkas, Konstantinos; Oliver, Stephen G. Hirotada4; Roumani, Ali2; Whitaker, Dawn R1 University of Cambridge, Department of Biochemistry, 1Purdue University, Biological Sciences, Wst Lafayette, IN, United Cambridge, United Kingdom States; 2Purdue University, Computer Science, Wst Lafayette, IN, United States; 3Purdue University, Chemical Engineering, Wst Objective: Large-scale approaches are fundamental in order Dedicated Lafayette, IN, United States; 4NAIST, Biological Sciences, Ikomo, to proceed from genome sequencing to systems biology. Posters Japan Quantitative screening in yeast allows the assessment of phenotypic effects due to the deletion of one of the two wild- We envision a future where a dynamic community of online type alleles in diploid cells. The information extraction from such resources (both information and computational resources) act genome-wide functional analyses across diverse environments cooperatively, with each resource focusing on its strength and is an essential step in the effort to build a eukaryotic whole-cell expertise, and linking to the strength and expertise of other model. resources. This vision of collaboration and sharing is part of a Results: Top-down analyses have been performed for suite of strategies that are collectively now referred to as Web population-profiling data of ca. 6000Saccharomyces cerevisiae 2.0. EcoliHub will not replace existing information resources; deletants grown in chemostats under 9 different experimental the goal is to add value to these resources by: 1) improving the conditions (C-, N- and P-limitation, in the presence of Cd, Cr, Cu, ability to share information and computational services among diamide, peroxide, and in grape-juice). By employing the yeast resources, 2) allowing resources to be combined (piped together) deletion collection of heterozygous mutants, the fitness profile of in new ways, without requiring additional development effort the entire yeast genome has been investigated. This holistic view by the provider, 3) improving the community’s ability to find of the yeast phenome led to the identification of numerous yeast information and resources, 4) providing new information and genes displaying haploinsufficiency/ haploproficiency. resources that ‘fill in the gaps’ between existing resources and Conclusions: The integration of these distinct population- improve the quality of information provided by all participating profiling data sets provides information, suggestions and evidence E. coli resources. Current status of the EcoliHub project will be which pave the way for the assessment of the contribution of all described. Funded by NIH GMS U24 GM077905 yeast genes to growth rate and fitness. Nevertheless, the real challenge is to elucidate why the growth rate of yeast is affected DS2-4-14 by these specific gene deletions under these conditions. Having identified classes of genes encoding proteins with, in the terms Moses: Developing wisdom for Microorganism systems of Metabolic Control Analysis, high flux-control coefficients, biology concerning energy and Saccharomyces cerevisiae an attempt is being made to model the pathways for protein Westerhoff, Hans V.1; Mensonides, Femke2; Kell, Douglas B.3; synthesis and protein turnover. Messiha, Hanan4; Kuchler, Karl5; Glaser, Walter5; Valachovic, Martin5; Reuss, Matthias6; Zakhartsev, Maksim6; Ruoff, Peter7; DS2-4-16 Woelfl, Stefan8; Kitanovic, Ana8; Nardelli, Maria1; Sharkey, Kieran1; Steuer, Ralf1; Verma, Malkhey1; Wilkinson, Steve1; Snoep, Jacky The beginning of the ends: A curvature-mediated L.1; Bakker, Barbara2 mechanism for localization of lipids to bacterial poles 1MCISB, Systems Biology, DTC, Manchester, United Kingdom; Huang, KC1; Mukhopadhyay, Ranjan2; Wingreen, Ned3 2NISB, Amsterdam, Netherlands; 3MCISB, Manchester, United 1Stanford University, Bioengineering, Stanford, California, United Kingdom; 4MCISB, Manchester, United Kingdom; 5Max F. Perutz States; 2Clark University, Physics, Worcester, Massachusetts, Laboratories, Vienna, Austria; 6Institute of Bioverfahrenstechnik, United States; 3Princeton University, Molecular Biology, Princeton, Stuttgart, Germany; 7University of Stavanger, Stavanger, Norway; New Jersey, United States 8IPMB, Heidelberg, Germany Objective: In the past decade, intracellular fluorescence Objectives: Develop a domino approach to Systems Biology of microscopy has fashioned a new appreciation for the diversity entire organisms of ways in which proteins organize and segregate on bacterial Clarify yeast energetics membranes. Though some targeting anchors are known, cellular Interrelate control analysis, regulation analysis and integral symmetry breaking ultimately requires molecular components that modelling self-organize. Develop methodology for the silicon human program including Results: We propose a novel equilibrium mechanism, based on model driven data handling the two-dimensional curvature of the membrane, for spontaneous Results: Yeast metabolism was split up into components using lipid targeting to the poles and division site of rod-shaped a bow-tie concept, with Gibbs energy of ATP hydrolysis (ATP) as bacterial cells. If one of the membrane components has a large nucleus. Kinetic models were made/collected of six surrounding intrinsic curvature, the geometrical constraint of the plasma subnetworks. These were each further optimized by one of membrane by the more rigid bacterial cell wall counteracts the

168 ICSB 2008 attractive interaction between like lipids and leads to microphase EPS (fuco-glucan) under all conditions studied. Since the reported separation. Using Monte Carlo simulations of our membrane monosaccharide composition of all EPS produced by Halomonas energetics model, we find that the resulting clusters of high- species usually contain glucose as the major component, it curvature lipids are large enough to spontaneously and stably was unclear which genetic mechanisms are responsible for the localize to the two cell poles and septal regions, and could have high fucose frequency of this new biopolymer. Considering the similar utility to lipid rafts as a stage for targeting proteins involved significance of fucose-rich polysaccharides in cell proliferation in a wide variety of biological processes. and aging, we were interested in elucidating the metabolic Conclusions: Evidence of localization of the phospholipid reactions and pathways that are important for the production of cardiolipin to the poles of bacterial cells suggests that protein this novel biopolymer. However, there is very limited information targeting may depend on the membrane’s heterogeneous about the mechanisms involved in the biosynthesis of EPS from lipid content. Recent experiments have also shown that the extremophiles and no report about a systematic approach to EPS osmoregulatory transporter ProP localizes to the poles by production by Halomonas sp. targeting cardiolipin in the rod-shaped bacterium Escherichia coli. Results:A genome-scale approach, integrating the available More generally, aggregates of lipids, proteins, and lipid-protein biochemical information, constitutes an important step toward complexes may localize in response to features of cell geometry understanding the polysaccharide production mechanisms in incapable of localizing individual molecules. Halomonas sp. and their possible improvement. Therefore, in this study, Chromohalobacter salexigens (formerly Halomonas DS2-4-17 elongata DSM 3043) was used as model organism. In this work, we present a genome-scale metabolic reconstruction (iOHYE617) Characterizing host-parasite interactions in leishmaniasis: for C. salexigens DSM43, which includes 1226 metabolic and A multi-cellular systems analysis transport reactions across 97 metabolic pathways. This model Chavali, Arvind; Papin, Jason was used to analyze the physiological capabilities of Halomonas University of Virginia, Biomedical Engineering, Charlottesville, VA, sp. during polysaccharide production. To study the physiological United States capacities in silico, an objective function was formulated to simulate EPS production. Objective: Systems-based approaches can facilitate the Conclusion:Overall, we present evidence supporting that Flux characterization of many components and interactions that Balance Analysis (FBA) of the reconstructed metabolic network constitute host-parasite relationships. In particular, the T-helper for Halomonas sp. provides results that are in agreement with

(TH) cell response following infection with Leishmania major (an physiological observations. Thus, as for other organisms, the agent of cutaneous leishmaniasis) in different strains of mice has reconstructed genome-scale metabolic network provides Posters been extensively studied. The production of TH subsets attributed an important framework which allows us to compare model Dedicated to phenotypes of resistance (TH1) or susceptibility (TH2) have predictions with experimental measurements and eventually been observed in various murine strains. Here, we applied an generate hypotheses to improve microbial polysaccharide agent-based modeling approach to capture the process of TH biosynthesis. differentiation in the lymph node of mice while accounting for interactions between T-cells and dendritic cells (DCs). DS2-4-19 Results: The systems-based model accounted for a heterogeneous population of TH cells including naïve T-cells, Quantitative studies of bacterial growth control 1 1 2 effector TH1s, effector TH2s, regulatory T-cells (TREG), non- Scott, Matthew ; Mateescu, Eddie ; Zhang, Zhongge ; Joseph, 3 1 polarized T-cells (TNP) and blast cells. An individual T-cell is Simpson ; Hwa, Terence allowed to integrate stimuli such as T-cell receptor (TCR) and 1University of California - San Diego, Center for Theoretical major histocompatibility complex (MHC)-peptide affinity, time Biological Physics, La Jolla, United States; 2University of of interaction with a DC and cytokines in the environment. The California - San Diego, Division of Biology, La Jolla, United States; aggregate of these stimuli determined the contribution of an 3University of California - San Diego, Dept of Chemistry and individual T-cell to the overall TH differentiation program. Biochemistry, La Jolla, United States Simulations were run for six-week periods, with initial conditions spanning various levels of antigen dose, TCR/MHC-peptide Objective: The goal of systems biology is to connect affinities and other stimuli. The outputs of the model, namely the molecular control strategies to the physiology of the organism. total numbers of TH1, TH2, TREG or TNP cells, for a given set of initial A fundamental aspect of bacterial physiology ripe for system- conditions were evaluated. The influence of cytokines such as level analysis is the growth laws: that irrespective of media, interleukin-12 (IL-12) or interleukin-4 (IL-4) in the model were also the macroscopic composition of the cell, e.g., the RNA, DNA explored. Removing IL-12 disrupted the production of a dominant and protein content, are functions of the growth rate alone. We

TH1 phenotype, while removing IL-4 did not qualitatively affect any have developed a simplified model that incorporates aspects simulations when compared to simulations with both IL-12 and of bacterial growth control with metabolism, and tested various IL-4 present. aspects of the model experimentally. Conclusions: Our model characterized emergent properties of Results: Our model highlights ribosome allocation as the key the TH differentiation process in response to antigenic stimulation. control mechanism, and uses flux balance to describe nutrient The immune response to L. major constitutes a multi-cellular uptake. The model is able to describe available data in the process and the actual pathological states of resistance and literature by a single fitting parameter. We then tested various susceptibility arise from an orchestration of interactions between components of the model by perturbing the ribosome function diverse populations of cell types and cytokines. in two ways: a) by globally retarding translation via antibiotic treatment, and b) preferentially retarding the translation of highly DS2-4-18 translated mRNA via a mutant 16S rRNA. Experiments were performed for various strains of E. coli, grown in media with Metabolic reconstruction and modeling of microbial different carbon and nitrogen sources that gave a 10-fold range in biopolymer production the growth rate of the wildtype cells. Quantitative measurements Ates, Ozlem1; Kazak, Hande2; Arga, Kazim Yalcin2; Toksoy Oner, were made to determine the ribosome and protein contents, as Ebru2 well as the fraction of disrupted ribosomes for each perturbation 1Marmara University, Chemical Engineering, Istanbul, Turkey; applied in each growth medium. Our model is able to predict 2Marmara University, Bioengineering, Istanbul, Turkey the nontrivial dependence of the growth rate on the applied perturbations for the different media, using a single fitting Objective: Microbially produced exopolysaccharides (EPS) act parameter which is based on the known biochemistry of the as new biomaterials with wide range of applications in many applied perturbations. industrial sectors. Recently, a newly isolated extremophile Conclusions: Understanding the global regulation underlying Halomonas sp. AAD6 was found to produce a rare fucose-rich growth physiology provides a platform for studies involving

ICSB 2008 169 nutrient limitation or synthetic circuits affecting growth. The considered; the buffering capacity is predicted to be minimum present investigation suggests molecular mechanisms necessary in the neutral range; it increases, as the pH gets alkaline or for robust control, independent of the growth medium, that acid. The effect of changing the phosphate pool into a fructose persist even in the presence of antibiotics or mutations designed 1,6-diphosphate pool during glycolysis is then investigated; the to inhibit ribosome function. model shows that the buffering capacity decreases significantly. Conclusions: The predicted buffering capacity qualitatively DS2-4-21 agrees with experimental measurements on cells: the buffering capacity is minimum at neutral pH (optimum pH of E. coli), Detailed 13C metabolic flux analysis of Saccharomyces and increases for decreasing or increasing pH. Phosphate and cerevisiae deletion mutants growing on glucose and glutamate are the main buffers, but other amino acids could be galactose reveals NADPH supplying function of malic considered in the future. From the variation of cytoplasmic pH and enzyme phosphate pool in anaerobic glycolysing E. coli, we use our model Schneider, Konstantin1; Yang, Tae Hoon2; Heinzle, Elmar1 to calculate the change in buffering capacity during glycolisis. We 1Saarland University, Biochemical Engineering Institute, plan to test this prediction with experiments in the near future. Saarbruecken, Germany; 2Univ. of Louisville, James Graham Brown Cancer Center & Dept. of Surge, Louisville, United States DS2-4-23

Malic enzyme is long known for its important role during growth of Modelling of methylglyoxal detoxification pathway in Saccharomyces cerevisiae on substrates like acetate and ethanol. Enteric Bacteria In early studies no phenotypic difference was observed between Almeida, Camila; Ozyamak, Ertan; Miller, Sam; Moura, the deletion strain BY4742 mae1∆ and the corresponding Alessandro; Booth, Ian; Grebogi, Celso wildtype strain BY4742 when observing parameters like substrate University of Aberdeen, Aberdeen, United Kingdom uptake, product secretion, specific growth and oxygen uptake rates (Velagapudi et al., 2006). A simplified 13C metabolic flux The objective of the project is the study of methylglyoxal (MG) analysis (MFA) during growth in microtiter plates on galactose detoxification and ionic homeostasis in Escherichia coli via and glucose and using MALDI-MS indicated significant mathematical and data based modelling. Dedicated differences (Velagapudi et al., 2007). The present study was MG is a toxic electrophile that reacts with electron rich centers of Posters designed to elucidate the detailed function of MAE1p applying DNA and proteins. It can diffuse freely across the membrane and continuous cultivation and a detailed isotopomer model including can also be produced endogenously when there is an imbalance mitochondrial compartimentation and reversible reactions. of sugar metabolism. Depending on its concentration, it inhibits All relevant substrates and products were measured using HPLC, cell growth or causes cell death. off-gas analysis and biomass determination. Labeling patters of The main detoxification pathway is glutathione (GSH) dependent proteinogenic amino acids derived from protein hydrolyzates and and involves two enzymes: glyoxalase I and glyoxalase II. Once extracted intracellular metabolites were determined using GC-MS. inside the cell, MG reacts with GSH, forming hemithiolacetal (HTA) In vivo flux estimation used biochemical reaction stoichiometry and two subsequent enzymatic reactions transform it first to and carbon isotopomer modeling of the entire reaction network S-lactoylglutathione (SLG) and then to D-lactate, the last product (Yang et al., 2008). The analysis allowed closing carbon and being non-toxic. electron balances to nearly 100% indicating the reliability of the The intermediate compound SLG, activates potassium channels results. The fluxes were almost identical between both strains (KefB and KefC), leading to a potassium efflux. To counterbalance when grown on glucose. During growth on galactose, however, this loss of cations, there is an uptake of protons, which lowers the pentose phosphate pathway activity was significantly larger in the internal pH and protects the cell, enhancing the chance of the mae1∆ strain. We conclude that MAEp plays an important role survival. in the NADPH supply of S. cerevisiae under respiratory growth. Experiments show that increasing SLG levels from a wild type References: Velagapudi, V. R., Wittmann, C., Lengauer, T., strain results in an increase in potassium efflux, indicating that the Talwar, P. and Heinzle, E. (2006). Process Biochem 41, 2170-79. system is in a state of sensitivity to changes in SLG levels. Velagapudi, V. R., Wittmann, C., Schneider, K. and Heinzle, E. Therefore there is an interest in estimating the levels of the (2007). J Biotechnol 132, 395-404. pathway’s intermediate compounds, specially SLG, and we make Yang, T.H., Frick, O., Heinzle, E. (2008) BMC Systems Biology, use of mathematical models to predict these levels. 2:29. The first proposed model does not include KefB or KefC activation. The strategy adopted involves the exchange of DS2-4-22 information between model and experiments, where the model guides new experiments that could then validate the model. Modelling the buffering capacity of bacterium E. coli The importance of pH on the detoxification pathway is being Richard, Morgiane1; Miller, Samantha2; Moura, Alessandro3; further investigated. The ultimate and challenging objective is to Booth, Ian2 model the system including both pH and ionic homeostasis, to 1University of Aberdeen, School of Medical Sciences / obtain a better understanding of cell responses when subjected Department of Physics, Aberdeen, United Kingdom; 2University to methylglyoxal stress. of Aberdeen, School of Medical Sciences, Aberdeen, United Kingdom; 3University of Aberdeen, Department of Physics, DS2-4-24 Aberdeen, United Kingdom Quorum signal integration in differentiating microcolonies Objectives: The buffering capacity of the cytoplasm is believed to Bischofs, Ilka1; Hug, Josh2; Liu, Aiwen3; Lee, David3; Wolf, be one of the means by which bacteria achieve pH homeostasis, Denise1; Arkin, Adam3 which is critical for microorganisms that may encounter wide 1Lawrence Berkeley Lab, Physical Biosciences Division, Berkeley, variations of environmental pH. However, experimental data on United States; 2UC Berkeley, Electrical Engineering, Berkeley, buffering capacity are limited. This project aims at developing a United States; 3UC Berkeley, Bioengineering, Berkeley, United mathematical model to predict the extent and change of buffering States capacity of the cytoplasm of bacteria Escherichia coli. Results: The model calculates the change in pH of a buffer Objective: RapA-phrA is a member of a complex quorum system solution induced by addition of strong acid. The concentration involved in regulating stress reponse coordination at the transition of protons after addition of strong acid can be expressed as a from vegetative growth to stationary phase in Bacillus subtilis. function of the concentrations and dissociation constants of the Here we study quorum signal integration both theoretically and buffers present in the solution, the volume of the solution and the experimentally. volume and concentration of the aliquot of strong acid. A buffer Results: By means of a computational model we find that the solution containing glutamic acid, phosphate and glutathione is particular feedback architecture of the central phosphorelay

170 ICSB 2008 supports a division operation of input kinase activity (encoding DS2-4-27 nutrition an other signals) with respect to the quorum modulated rap signals. This suggests that cells might normalize Prediction of metabolic strategies by optimization environmental signals to the population count. Using time lapse techniques microscopy we find that the RapA/PhrA operon is induced Teusink, Bas1; Wiersma, Anne1; Notebaart, Richard2; Smid, Eddy1; heterogeneously in sporulating microcolonies. RapA off cells exit Molenaar, Douwe1 the cell cycle at the onset of stationary phase to sporulate while 1TI Food and Nutrition, NIZO food research, Ede, Netherlands; RapA on cells continue to divide and delay sporulation. Since 2Radboud University Nijmegen Medical Center, Centre for the RapA phosphatase and its “quorum sensing” inhibitor are Molecular and Biomolecular Informatics, Nijmegen, Netherlands expressed simultaneously only in the dividing cell population, the previous result implies that this subpopulation is monitoring its Objective. We have built several genome-scale metabolic models size allowing for a normalized food count on the basis of cells for a number of food-related bacteria. These models are based actively competing for nutrients only. on bioinformatics, literature, and experimental evidence for the Conclusions: We suggest that the various rap-phr systems activity of specific pathways. They are important tools in the lab might serve cell-cell communication among differentiated for data analysis, integration and visualization. Predictive modeling subpopulations in, for example, biofilms, where environmental has also been performed with genome-scale metabolic models. signals need to be interpreted based on the spatio-temporal Optimization of an objective function –referred to as flux balance community structure. analysis (FBA)- has been often used to predict flux distributions in metabolic networks, but it fails miserably in lactic acid bacteria. DS2-4-26 The reason is that FBA optimizes yields and hence assumes metabolic efficiency as the best strategy for fitness, whereas lactic Industrial systems biology of Saccharomyces cerevisiae: acid bacteria are highly inefficient consumers of sugars. Here Succinic acid production we report two approaches to move forward. Results. First, we Otero, Jose Manuel; Olsson, Lisbeth; Nielsen, Jens hypothesized that poor selection conditions may be chosen such Chalmers University of Technology, Systems Biology, Dept. of that efficiency is an important strategy for fitness improvement Chem. & Biological Eng., Gothenburg, Sweden even for inefficient lactic acid bacteria. Specifically, we were able to adapt Lactobacillus plantarum to growth on glycerol as main Objective: Saccharomyces cerevisiae is a proven, robust, carbon source, an unusual and poor carbon source for these industrial production platform used for expression of a wide range bacteria. Under these conditions FBA with biomass production of therapeutic agents, added-value chemicals, and commodities. as objective was able to predict product formations. This study

S. cerevisiae is also a eukaryotic platform upon which systems thus provides an example how the conditions can be steered Posters biology tools such as genomics, transcriptomics, metabolomics, to fulfill the assumption of the modeling approach. Second, we Dedicated fluxomics, proteomics and bioinformatics have been developed. reconsidered growth rate as the most general proxy for fitness. This powerful multi-disciplinary and integrative approach has Consequently, the stoichiometric modeling approach (yields only) enabled metabolic engineering for overproduction of added-value was changed to a kinetic approach. We therefore constructed a chemicals such as succinic acid. Succinic acid is a key building core model of balanced bacterial growth and used optimization block molecule for further conversion to precursor molecules such of growth rate to predict metabolic strategies. Most importantly, as tetrahydrofuran, 1,4-butanediol and butyrolactone, with over we included costs associated with protein synthesis. This 160 million kgs currently produced petro chemically with a sale analysis provided a cost-benefit analysis explaining inefficient price of $US 5-9/kg. While there are several prokaryotic platforms metabolism as a means to grow as fast as possible. Conclusions. for succinic acid production already in development, S. cerevisiae Optimisation techniques can be useful in predicting optimal offers distinct process advantages. S. cerevisiae has been metabolic strategies if underlying assumptions and simplifications awarded GRAS status for use in human consumables, grows well are critically considered in the light of biological relevance. at low pH significantly minimizing purification and acidification costs associated with organic acid production, and can utilize DS2-4-28 diverse carbon substrates in chemically defined medium. The objective of this work is to apply systems biology approaches to Mathematical modeling of the regulatory action of metabolic engineering for transformation of S. cerevisiae into a antisense RNA IsrR in Synechocystis PCC 6803 microbial cell factory. Rho, Seong-Hwan; Jansen, Stefan; Schoen, Verena; Ude, Results: Here we describe the use of systems biology tools to Susanne; Timmer, Jens; Hess, Wolfgang R. drive C6 carbon flux to succinic acid by enhancement of the University of Freiburg, Freiburg, Germany two native pathways for succinic acid production: the TCA and glyoxylate cycles. S. cerevisiae does not naturally accumulate Objective: Antisense RNAs are participating in diverse cellular succinic acid, yet through the use of in silico metabolic predictions processes in various ways such as degradation, stabilization, or guiding targeted gene deletions and overexpression, mutants that transcriptional interference of their target mRNAs. In the simple overproduce succinic acid have been engineered and thoroughly cyanobacterial model Synechocystis PCC6803, several hundreds characterized. of cis-antisense RNAs have been predicted and need to be Conclusions: In silico predictions based on the reconstructed characterized. In this study, we focus on IsrR, a 177-nt antisense genome-scale metabolic network provide non-intuitive gene RNA transcribed from the non-coding strand of the isiA gene. isiA targets that coupled with the integration of physiological encodes IsiA, a 37 kDa protein homologous to the Photosystem II characterization and transcriptomics have resulted in succinic chlorophyll-binding protein CP43. acid overproduction. Further engineering is required to reach Results: The expression of IsiA shows a robust pattern during the industrially relevant yields, but proof of concept has been long-term growth and is sensitive to iron limitation and oxidative established. stress. Under the iron depleted culture condition, the expression of IsiA increases, while it is at the basal level under the normal condition. This robust nature of IsiA expression seems to be an outcome of interplay of IsrR and other cellular processes yet to know. A data-based mathematical modeling approach was taken in a bottom-up fashion to elucidate the regulatory mechanism of isiA. Models with different regulatory schemes are compared with the quantitative time series data such as the activity of each promoter monitored by expression of the reporter gene luxAB. Conclusions: Our study suggests that a simple co-degradation model cannot explain the complex expression pattern of IsiA. Introduction of the control on the repressor Fur and translational

ICSB 2008 171 regulation dependent on the iron level is crucial. This study will well as phosphate conservation, was required to properly predict contribute to learn one of possible regulatory schemes of small the quantitative dynamics of intra- and extracellular metabolites, antisense RNAs in cyanobacteria. ADP/ATP, NAD+/NADH and inorganic phosphate levels as observed by in vivo 13C NMR data. In addition, simulations DS2-4-29 of glycolysis restart after a period of starvation revealed the potential use of the mixed-acid branch in providing ATP to initiate Alginate synthesis in Pseudomonas fluorescens as a model glycolysis. system for global metabolic network studies Conclusions: Based on a kinetic modeling approach, we Valla, Svein identified key allosteric regulation mechanisms needed to properly Norwegian University of Science and Technology, Department of describe experimentally determined metabolite dynamics, as well Biotechnology, Trondheim, Norway as glucose uptake and product formation in glycolysis of L. lactis. This knowledge will be used to compare the other two lactic acid- Objective: To use alginate synthesis in the bacterium producing bacteria studied in this project. Pseudomonas fluorescens as a model system to analyze metabolic network dynamics and robustness. The project is part DS2-4-31 of a joint European initiative on Microbial Systems Biology (SysMo), started spring 2007. Quantitative analysis of triacylglycerol mobilization Results: Alginate is in bacteria an extra-cellular polysaccharide Natter, Klaus1; Zanghellini, Juergen2; Jungreuthmayer, Christian3; that is composed of two different sugar moieties, mannuronic Thalhammer, Armin2; Kurat, Christoph F.1; Kohlwein, Sepp D.1; acid and guluronic acid. In genetically engineered strains of P. von Gruenberg, Hans-Hennig2 fluorescens the synthesis of the polymer could be controlled over 1University Graz, Inst. of Molecular Biosciences, Graz, Austria; a wide range of levels by an externally added inducer in cells 2University Graz, Institute of Chemistry, Graz, Austria; 3Trinity cultivated in chemostats. The alginate biosynthesis genes are College, Trinity Center of Bioengineering, Dublin, Ireland clustered and we control their expression from the heteorologous inducible Pm promoter. This promoter can be activated in a Objective: Triacylglycerol (TAG) is an important storage graded way such that alginate synthesis can also be set over compound in eukaryotic cells, stockpiled together with Dedicated a very wide range of levels. At full induction around 40% of the other neutral lipids in cytosolic lipid droplets. We aim at the Posters sugars consumed are channeled into polymer synthesis. Samples understanding of the importance and mechanisms of TAG have been collected from off and on states and analyzed at the metabolism in the yeast S. cerevisiae. In this study, lipid transcriptome, proteome and metabolome levels. In addition an in mobilization in stationary phase yeast cells reentering vegetative silico model of whole cell metabolism has been constructed, and a growth was quantitatively analyzed using a combined theoretical genome-saturated transposon insertion library has been screened and experimental approach. Cellular growth and size, glucose with respect to the ability of the mutants to synthesize alginate. At uptake and ethanol secretion were measured as a function of time the time of writing all these data are being processed and will be and used as input for a dynamic flux-balance model. integrated into the in silico model aiming at developing a predictive Results: By combining dynamic mass balances for key model for metabolic network behavior in P. fluorescens. metabolites with a detailed steady-state analysis, we used Conclusions: Alginate synthesis in P. fluorescens can be experimental data to iteratively train our model network controlled and stably maintained at different levels and at steady and calculate the time dependent degradation of cellular state in chemostats. Analyses of ‘omics samples have provided triacylglycerol. We found that during pre-logarithmic growth TAG a broad picture of the intracellular effects of turning alginate is used as a feedstock for membrane lipids. Moreover we show synthesis on. The results of analysis of the correlation between that TAG degradation is proportional to the rate of change of the transposon insertion points and the capacity to synthesize alginate cell’s surface area. Our predictions are in excellent agreement with indicate that many non-obvious genes interfere with alginate experimental data, both, qualitatively as well as quantitatively. biosynthesis. Conclusions: We identified lipolysis as a key process during the transition of quiescent cells into the proliferating state DS2-4-30 and propose that by observing the cellular surface area conclusions on the lipolytic activity can be drawn. Methodically Kinetic model of Lactococcus lactis glycolysis: Refined we demonstrated that our approach is able to simultaneously control by allosteric regulation describe the time evolution of various key metabolites in a Musters, Mark1; de Vos, Willem M.1; Teusink, Bas2 consistent and self-contained manner. 1Wageningen University, Laboratory of Microbiology, Wageningen, Netherlands; 2Top Institute Food and Nutrition (WCFS), DS2-4-32 Wageningen, Netherlands PRODORIC - a platform for the elucidation of gene Objective: Within the SysMo project “Comparative Systems regulatory networks in prokaryotes Biology of Lactic Acid Bacteria”, the aim is to compare the Klein, Johannes; Retter, Ida; Leupold, Stefan; Muench, Richard glycolysis of three related, but functionally different, lactic Technical University Braunschweig, Microbiology, Braunschweig, acid–producing bacteria i.e., Lactococcus lactis, Streptococcus Germany pyogenes and Enterococcus faecalis. Of these, the regulation of glycolysis in L. lactis has been studied for decades, but many Objective: Although the genomes of several hundreds of aspects remain elusive. Specifically the mechanism for the bacterial species are entirely sequenced, the elucidation of the switch between homolactic to mixed-acid fermentation has been corresponding gene regulatory networks is still a challenging the subject of many debates in the literature. Unlike previously process that requires extensive analyses on multiple levels. We reported models, we developed a nonlinear kinetic model of L. present a platform for the retrieval, prediction and visualization lactis glycolysis in which allosteric regulation, conservation of the of gene regulatory networks in prokaryotes that enables an phosphate pool and the mixed-acid branch were all implemented integrated and interactive analysis. The framework consists of the in the convenience kinetics framework. PRODORIC database, the Virtual Footprint prediction tool and the Results: Our model is a revision of the Hoefnagel model ProdoNet visualization tool. (Hoefnagel et al., 2002, Microbiology 148: 1003 – 1013) in which Results: PRODORIC, the Prokaryotic Database Of Gene careful examination of the literature resulted in initial estimates Regulation, provides a comprehensive source of manually curated of the parameter values. The kinetic model was subsequently knowledge on regulation of gene expression in prokaryotes. The fitted on published experimental time-series data. The allosteric database provides integrated data of transcriptional regulators inhibition of the PTS system was incorporated to describe the and their corresponding binding sites, gene expression patterns control of glucose uptake when excessive amounts of glucose are and related information. The pattern matching tool Virtual available. Incorporation of other allosteric regulation systems, as Footprint complements the data obtained from experimental

172 ICSB 2008 evidence by the prediction of regulatory interactions and whole Results: We have screened about 4000 transposon insertion regulons in bacterial genomes. For the mapping of prokaryotic mutants for their alginate production and identified about 400 that genes and corresponding proteins to common gene regulatory produce significantly different amounts of alginate when grown and metabolic networks, ProdoNet provides an intuitive tool for in micro titer plates. These candidates are then retested in shake the visualization of experimental and predicted data. flasks, and the gene affected by the transposon is identified by Conclusions: The derived data is used for different approaches sequencing. The transposon used contains a promoter-less to create models for prokaryotic gene regulatory networks. reporter gene allowing us to screen the same library for genes Thus, in a process of data mining, prediction, visualization being expressed during encystment, and several candidate and modeling, the flow of data is turned into information and genes have been identified. Some of the genes identified as being knowledge. The PRODORIC platform is accessible at http://www. turned on during encystment have been shown to affect alginate prodoric.de. production or composition. Conclusions: The data from screening the transposon library DS2-4-33 has identified genes that are necessary for alginate synthesis and others that have a negative impact on alginate production. Genes Evaluation phylogenetic diversity of biosurfactant that putatively could be involved in the biosynthesis of other cyst producing bacteria components have also been identified. These data will be valuable Grebenisan, Irina1; Olteanu, Violeta2; Cimpeanu, Carmen2; Alexe, in understanding more of the regulation of alginate production and Roxana3; Campeanu, Gheorghe2 encystment in A. vinelandii, especially when combined with results 1Faculty of Environmental Engineering, Environmental Sciences, from ongoing microarray studies. Bucharest, Romania; 2Faculty of Biotechnology, Bucharest, Romania; 3Faculty of Environmental Engineering and Land DS2-4-36 Reclamation, Bucharest, Romania Growth-rate dependent effects on bacterial gene Objective Biosurfactants have become an important product expression of biotechnology. They can be used in various industrial Klumpp, Stefan1; Lee, Pohan1; Liu, Xin1; Zhang, Zhongge2; Hwa, sectors such as environmental control and management, Terence1 agrochemicals and fertilizers, food and beverages, cosmetics 1University of California at San Diego, Center for Theoretical and pharmaceuticals, petroleum and petrochemicals, organic Biological Physics, La Jolla, CA, United States; 2University of chemicals, mining and metallurgy. They can be used as California at San Diego, Division of Biological Sciences, La Jolla, emulsifiers, de-emulsifiers, wetting agents, spreading agents, CA, United States foaming agents, functional food ingredients and detergents. The Posters present work is an initial attempt to systematically screen for Objective: For fast growing bacteria, changes in gene expression Dedicated biosurfactant-producing microorganisms and to evaluate their are often accompanied by changes in growth rates. Because the phylogenetic diversity. macroscopic composition of bacteria (e.g., cell size, ribosome Results Ten soil and water samples were collected from both concentration, gene copy number) is known to vary greatly at undisturbed sites and contaminated sites (i.e., with organics and/ different growth rates, significant changes in gene expression or hydrocarbons) in the southeast Romania and Black Sea. The may arise “passively” just due to the growth rate change alone. soil and water samples were screened for biosurfactant producers Therefore, quantitative understanding of gene regulation requires by using a combination of cultural and molecular methods - a quantitative understanding of these passive effects. Towards FISH technique. The 505 colonies obtained were screened this end, we performed a joint theoretical/experimental study for for biosurfactant production in MSM - mineral salts medium E. coli. We characterized the growth-rate dependent expression containing 2% glucose. Forty-five of the isolates were positive for of “constitutive” genes, and explored regulatory strategies to biosurfactant production, representing most of the water and soil make gene expression independent of growth rates. tested. Putative biosurfactant-producing isolates were cultured Results: We analyzed quantitatively available data for various from the various soils and water and grouped using fluorescent in macroscopic parameters affecting gene expression in E. coli, situ hibridization analysis. and predicted the growth rate dependent expression for various Conclusions A phylogenetic tree was constructed by combining simple genetic circuits. We tested these predictions by driving the results of the present study with a survey of biosurfactant- the expression of several exemplary genes by various promoters, producing eubacteria and archaea found in the literature to for several strains of E. coli grown in a variety of media. For examine the diversity of biosurfactant producers. each strain and growth medium, mRNA levels were measured by qPCR and protein levels by activity assays or quantitative DS2-4-34 Western blotting. For a constitutively expressed gene, the mRNA level was found to be weakly growth rate dependent, while the Identification of genes involved in alginate production and expressed protein concentration decreased nearly inversely with encystment in Azotobacter vinelandii the growth rate, in accordance with predictions of our model. Ertesvåg, Helga1; Mærk, Mali1; Jakobsen, Øyvind M.2; Vie, Ane Weak growth-rate dependence was predicted and observed also Kjersti1; Sletta, Håvard2 for autorepressing genes and for genes under negative control by 1Norwegian University of Science and Technology, Department an autorepressor. of Biotechnology, Trondheim, Norway; 2SINTEF, Materials and Conclusions: Our studies demonstrate that growth rate has Chemistry, Department of Biotechnology, Trondheim, Norway important and substantial effects on gene expression. These effects must be taken into account when analyzing gene Objective: Alginate is a linear copolymer of mannuronic and expression data under different condition. Buffering against these guluronic acid that is produced by brown algae and some effects may be an important requirement underlying the robust bacteria belonging to the genera Pseudomonas and Azotobacter. operation of endogenous genetic circuits in these bacteria, and Both genera use alginate as part of their extracellular slime should be a prime factor to consider in the design of robust capsule. The nitrogen-fixing soil bacteriumAzotobacter vinelandii synthetic circuits. is able to form desiccation resistant cysts under adverse growth conditions. The outer layer of these cysts contains a layer of alginate rich in the gelforming guluronic acid residues, while alginates in the inner layer contains more mannuronic acid. Cyst coats also contain lipids, other carbohydrates and proteins, however, knowledge on the composition of the cysts and the regulation of the encystment process is very limited. In this study we aim at identifying genes and pathways that are involved in biosynthesis of alginate and encystment in A. vinelandii.

ICSB 2008 173 DS2-4-38 DS2-4-40

Bioinformatics to improve biotechnological enzyme and Analysis of genetic diversity among Korean xanthomonas metabolite production oryzae pathovar oryzae strains Heijne, Wilbert; Roubos, Hans; Pel, Herman Lee, Byoung-Moo1; Park, Young-Jin1; Song, Eun-Sung1; Kim, DSM Food Specialties, R&D Genetics, Delft, Netherlands Jeong-Gu1; Cho, Hee-Jung1; Noh, Tae-Hwan2 1National Institute of Agricultural and Biotechnology, RDA, Suwon, Objective: A wide variety of biotechnological fermentation Republic of Korea; 2Honam Agricultural Research Institute, RDA, processes exploit microorganisms for the production of proteins Iksan, Republic of Korea such as (food) enzymes and metabolites such as antibiotics. At DSM, bioinformatics approaches and genomics technologies Xanthomonas oryzae pv. oryzae (Xoo) causes bacterial blight (BB) are fully integrated in our research to develop high producing in rice. The strains of Xoo have been classified into many races microbial strains. based on their virulence. Each country has its own differential Results: For example, we sequenced Aspergillus niger CBS system for race differentiation. Therefore, the differential cultivars 513.88, which is a key organism for the production of, primarily, used for race differentiation are different depending on the enzymes. Based on the genome sequence, transcriptomics, country. Korean X. oryzae pv. oryzae were classified into five proteomics and metabolomics were established to provide races based on their virulence. In this study, the races of Korean new insights in the physiology of the organism. Our program Xoo based on their virulence were somewhat different from for rational improvement of protein production implicates their RFLP patterns. Therefore, we redefined Korean Xoo races improvement of the generic host strain (bug), the gene by differences in RFLP patterns using repetitive DNAs and avr construct (plug) and the fermentation process. We, for example, gene as a probe. RFLPs using repetitive DNAs and avr genes systematically studied codon and codon-pair bias in the genomes were useful tools for race diagnosis and classification of Korean of our preferred host organisms, including the fungi Aspergillus Xanthomonas oryzae pv. oryzae. niger and prokaryotes E.coli and B.subtilis. Conclusions: Based on the observed bias, computer algorithms DS2-4-41 were developed to optimize codon usage obtaining more Dedicated efficiently translated gene sequences. The developments in Physiological analysis and gene expression profiling of Posters synthetic biology enable to rapidly obtain these optimized the pigment-deficient mutants in xanthomonas oryzae constructs and test them in industrially relevant strains. pathovar oryzae Park, Young-Jin1; Song, Eun-Sung1; Noh, Tae-Hwan2; Kim, DS2-4-39 Hyungtae3; Lee, Byoung-Moo1 1National Institute of Agricultural Biotecnology, RDA, Suwon, Cooperation and competition in the secretion of invertase Republic of Korea; 2Honam Agricultural Research Institute, RDA, by Saccharomyces cerevisiae Iksan, Republic of Korea; 3Macrogen Inc., Seoul, Republic of Schroeter, Anja1; Ruppin, Eytan2; Brenner, Naama3; Kreft, Jan- Korea Ulrich4; Schuster, Stefan1 1Friedrich-Schiller-University Jena, Department of Bioinformatics, Xanthomonas oryzae pv. oryzae (Xoo) causes bacterial blight (BB) Jena, Germany; 2Tel Aviv University, School of Computer Sciences of rice. A random insertional mutant library of Xoo KACC10331 & School of Medicine, Tel Aviv, Israel; 3Technion – Israel Institute was constructed using a Tn5-derived transposon, and the of Technology, Department of Chemical Engineering, Haifa, virulence of the mutants against the susceptible rice cultivar IR24 Israel; 4University of Birmingham, Centre for Systems Biology, was assayed. After the virulence assay, the M6 (aroE::Tn5) mutant Birmingham, United Kingdom that had reduced virulence against the rice plants was isolated. Furthermore, the M6 mutant had reduced pigment production. Saccharomyces cerevisiae growing on sucrose hydrolyses this In addition, we isolated another pigment-deficient mutant, M11 disaccharide into glucose and fructose outside of the cytoplasm. (aroK::Tn5), retained the virulence against the rice plants. Thermal The corresponding secreted enzyme - invertase - is mainly stored asymmetric interlaced-polymerase chain reaction (TAIL-PCR) and attached to the cell wall, with the remainder diffusing into the sequence analysis of the M6 and M11 mutants revealed that the medium. Its secretion can be regarded as cooperative behaviour. transposon was inserted into the aroE gene (encodes a shikimate The production of the enzyme is costly to the respective yeast dehydroganse) and the aroK gene (encodes a shikimate kinase) cells but is beneficial also to neighbouring cells as they may respectively. Furthermore, we used DNA microarrays to analyze take up the glucose generated. The genes coding for invertase gene expression profiles of the mutants. form the SUC multigene family. Interestingly, some S. cerevisiae cells secrete less invertase or none at all because some or all DS2-4-42 of the corresponding SUC genes are silenced. These cells can be considered to be cheaters which follow a selfish strategy Stochastic modelling of survival of E. coli exposed to that saves the metabolic effort to synthesize the enzyme but still methylglyoxal stress benefits from the glucose resulting from the action of invertase Karschau, Jens1; de Almeida, Camila2; Richard, Morgiane2; secreted by other cells. Greig and Travisano (Proc. R. Soc. Lond. Miller, Samantha3; Booth, Ian R.3; Kremling, Andreas4; de Moura, B 271 (2004) S25-S26) have performed experiments with mixed Alessandro2 populations of S. cerevisiae possessing either functional or 1University of Aberdeen, Molecular and Cell Biology, Aberdeen, deleted SUC2 genes. They determined the relative fitness (growth United Kingdom; 2University of Aberdeen, Physics, Aberdeen, advantage) of the defecting cells in dependence on the initial cell United Kingdom; 3University of Aberdeen, Molecular and Cell density. We present a game-theory oriented model analyzing their Biology, Aberdeen, United Kingdom; 4Max-Planck-Institute for results. In this model, the spatial heterogeneity of the glucose Dynamics of Complex Technical Systems, Systems Biology, concentration is taken into account, as it is a critical variable in Magdeburg, Germany correctly interpreting the pertaining experimental results. We make the plausible assumption that the glucose level near defector Objectives: The objectives of this project are to model the death cells is lower than near cooperative cells because the former take of E. coli caused by methylglyoxal induced double-strand breaks. up glucose without contributing to its formation. This implies an The assumptions are: incentive for invertase secretion to the yeast cells and, thus, for • Growth is inhibited for methylglyoxal concentrations (this is cooperation. known to be the case for concentrations higher than 0.3mM). • The only repair process taking place is that due to DNA repair enzymes; other repair processes and the detoxification of methylglyoxal are not included. • Methylglyoxal attacks guanine bases, which are assumed to be

174 ICSB 2008 evenly distributed on the DNA DS2-4-44 • Death is due to two closely opposed single-strand breaks forming a double-strand break. Breaks occur at the time of repair Ab initio regulatory element detection in nucleotide when endonuclease cleaves the DNA. sequences • The number of repair enzymes is assumed to be constant. Kelly, Steven1; Wickstead, Bill2; Gull, Keith2 The average death rate for E. coli K-12 as a result of 1University of Oxford, Oxford, United Kingdom; 2University of methylglyoxal exposure was calculated from the model, both Oxford, Sir William Dunn School of Pathology, Oxford, United analytically and by numerical simulation, as well as measured Kingdom experimentally. The E. coli strain used in the experiments lacks the main methylglyoxal detoxification pathway. Objective: The problem of ab initio regulatory element detection Results: The analysis and computer simulations of the model in nucleotide sequences is one of the biggest challenges in show that the death rate of cells is constant if the number of understanding biological diversity. Many eukaryotic organisms repair enzymes is large enough; increasing their number does not regulate their gene expression in surprisingly different ways. affect the death rate, which seems counterintuitive. Furthermore, One example of this are the trypanosomatid parasites, which the model predicts that the death rate is proportional to the regulate their gene expression entirely post transcriptionally. square of the methylglyoxal concentration; this is confirmed by Given this unconventional method of regulation, and also an our preliminary experimental results. unconventional set of transcription components encoded within Conclusions: For low enzyme numbers, the increase in number its genome, one might not expect to find the conventional of repair enzymes increases the probability that two damaged types of regulatory element in trypanosomatids. Indeed, bases that are sufficiently close together will be cleaved at the overrepresented sequence motifs, transcription factor binding- same time; therefore the model predicts that cells die more sites and other conventional regulatory elements have not yet quickly. Further experimental results are needed to check this. been reported in trypanosomatids. Results: We undertook to see Also the predicted dependence of the death rate on methylglyoxal whether utilising a sophisticated test for nucleotide compositional concentration needs to be better tested. segments could reveal information about the location and/or composition of regulatory sequence elements in trypanosomatids. DS2-4-43 We developed a novel program to evaluate changes in sequence composition using Bayesian segmentation, and employed this Masking of resistance mutations in efflux pump deficient program in an effort to identify potential regulatory elements in bacteria the trypanosomatid genome. This analysis has identified several Fange, David; Ehrenberg, Måns conserved regions of the upstream and downstream intergenic

Uppsala University, Dept. of Cell & Molecular Biology, Uppsala, regions of the ubiquitously expressed monotypic and polytypic Posters Sweden gene arrays in trypanosomatids, some of which are removed Dedicated during RNA processing and hence do not contribute to post- Objective: Antibiotic drugs are continuously being rendered transcriptional regulation. ineffective due to the emergence of resistance mutations, which Conclusions: We present here a novel method for identifying may affect drug-target binding sites, drug efflux pumps, or regulatory elements in nucleic acid sequence and identify several drug membrane permeation. We wished to investigate how the putative elements conserved in ubiquitously expressed genes. We growth-rate advantage due to resistance mutations in drug-target propose that some of these elements act to aid processing of the binding is modulated by the rates of active efflux drug pumping. transcript and that others act concertively by producing mRNA The rates of efflux pumps are especially interesting due to the secondary structure which modulates the stability of the mature efflux pump inhibitors that are currently being investigated as a messenger RNA. Using this process we also identify putative response to efflux dependent multi drug resistant pathogens. promoter sites and transcription termination sites and interrogate Results: We have constructed a model, which accounts for their distribution on a genome wide scale. passive drug influx and efflux by membrane diffusion, active drug efflux, and growth-inhibitory drug binding to intracellular DS2-4-45 targets. We describe the steady state conditions under which the model displays growth bi-stability, where the pathogen switches High-throughput phenotypic fingerprinting of from a state of rapid to a state of slow growth at a saddle-node saccharomyces cerevisiae deletion strains by Fourier bifurcation. The criterion for bi-stability is that the rate of drug transform infrared spectroscopy target binding needs to be fast in comparison to the drug efflux Böcker, Ulrike1; Martens, Harald2; Omholt, Stig W.3; Warringer, and drug unbinding. We also show that when drug efflux is slow Jonas4; Blomberg, Anders4; Kohler, Achim2 and antibiotic target binding is strong, target binding resistance 1Matforsk AS & CIGENE – Center for Integrative Genetics, mutations will neither affect the growth-rate in the rapid growth Norwegian University of Life Sciences, Ås, Norway; 2Matforsk regime nor the minimal inhibitory drug concentration. In the AS & CIGENE & IMT, Norwegian University of Life Sciences, Ås, slow growth regime, in contrast, target resistance mutations will Norway; 3CIGENE – Center for Integrative Genetics, Norwegian stimulate growth in relation to wild type. We show, finally, how the University of Life Sciences, Ås, Norway; 4Göteborg University, steady state analysis relates to experimentally relevant pre-steady Department of Cell and Molecular Biology, Göteborg, Sweden state cases. Conclusions: Drug efflux pump deficiency will not only make Objective: State-of-the-art high-throughput growth bacterial pathogens more susceptible to antibiotic treatment, but phenotyping of yeasts is based on single-channel optical may also mask the growth-stimulation effect of target resistance density measurements. Thereby the growth features of yeast mutations, seen under conditions of efflux pump proficiency. We knockout mutants can easily be compared to their wildtype. predict the rate of evolution of drug resistance be greatly reduced While the growth phenoptyping has been developed to a highly under conditions of drug efflux pump deficiency, which may be reproducible and high-throughput technique to determine ‘fitness’ clinically relevant. parameters for deletion strains, it is now desirable to go further and develop techniques that can measure genome-wide chemical fingerprints of metabolic products. Results: In this paper we propose Fourier transform infrared spectroscopy as a high-throughput technique for measuring chemical fingerprints of metabolic products of yeast deletion strains. Fourier Transform infrared (FTIR) spectroscopy is an emerging technique for the characterization and identification of cell and tissue components in intact cells. We have developed a protocol for combining the high-throughput instrumentation used for measuring fitness parameters with a high-throughput FTIR

ICSB 2008 175 measurement principle, where 384 sample wells are used for changes with time is not clear. Here we systematically studied spectroscopic measurements. Eighty Saccharomyces cerevisiae the time dependent dynamics of gene expression of enhanced BY4743 knockout mutants lacking genes related to fatty acid green fluorescence protein from different simple unregulated metabolism and having a detectable growth phenotype were and regulated synthetic gene systems using flow cytometry. analyzed by FTIR spectroscopy according to the new protocol. Contrary to the assumption, we found that there was no steady Our results show that the method is highly reproducible. Several state in the gene expression level for up to 12 hours. Both the of the examined deletion strains showed detectable FTIR mean gene expression level and variability among a genetically phenotypes in fatty acids and polysaccharides. identical population change as a function of time even in the Conclusions: The proposed method opens new possibilities exponential growth phase for both regulated and unregulated for the annotation of genes in genome-wide phenotyping, since systems. We also showed that the time dependent behaviors chemical fingerprints of products of metabolic pathways are of unregulated and regulated gene expression are different. A obtained in addition to the well established fitness parameters. simple mathematical model was formulated which captures the qualitative features of time dependent dynamics where gene Dedicated session 3-1: expression is coupled with the cell growth and division. Cell-to-cell variation DS3-1-12

DS3-1-10 Optical tweezers combined with a microfluidic device for studies of stress response in single cells Operator state dynamics in gene expression under Eriksson, Emma1; Sott, Kristin1; Petelenz, Elzbieta2; Enger, Jonas1; feedback Goksör, Mattias1 Pulkkinen, Otto; Berg, Johannes 1University of Gothenburg, Department of Physics, Gothenburg, University of Cologne, Institute for theoretical physics, Cologne, Sweden; 2University of Gothenburg, Department of Cell and Germany Molecular Biology, Gothenburg, Sweden

Objective: The dynamics of the operator state is an important Objective: Traditionally biological experiments are performed Dedicated source of fluctuations in prokaryotic gene expression. An example on large populations of cells, where measurements only reflect Posters is the repressed Lac system in E. coli extensively researched the average behavior of many cells at a certain point in time. In in the experiments of Yu et al. [1] and many others. The Lac the averaging of the result, information about the heterogeneity, repressor detaches from the DNA only once or twice within a stochastic behavior or synchronization of signaling pathways gets cell cycle, and these events are followed by a burst of protein lost. Even genetically identical cells show noise, which could be production. The purpose of this talk is to show how the dynamic important in the interpretation of cellular behavior. Lack of control description of the operator state changes when feedback is of the cellular microenvironment might also contribute to the introduced. In the Lac system, feedback is known to exist through broadening of population distributions. To reveal the heterogeneity the mechanism of autoactivation. among cells there is a need for techniques able to provide data Results: In the presence of feedback, the distribution of times on a single cell level with high temporal resolution and precise during which the gene is transcribed shows a clear deviation from environmental control. the naive mixture of two exponential distributions corresponding Results: By combining optical tweezers and microfluidic systems, to repressed and induced states. This distribution of these times we have developed a platform where we rapidly can change is found to be reflected in the distribution of expression levels the microenvironment around single cells. Two complementary across a cell population, and we compare our results to data from designs of this system are presented, one suitable for studies single-cell experiments in a population of cells. of fast biological processes, where the environmental change Conclusions: The mathematical analysis is carried out using is performed on a subsecond timescale, and one more suitable a piecewise deterministic model that retains operator state for studies over longer periods of time, where the complete fluctuations as the only source of noise. This approximation is environmental change is accomplished in less than 10s. The shown to capture the dynamic features of more complicated experimental platform is demonstrated on S. cerevisiae, where models, which feature translation bursts, up to moderate levels we monitor the response in various signalling pathways, upon of induction, demonstrating that the operator state dynamics is environmental stress (glucose starvation, oxidative stress and robust against small-scale fluctuations. Furthermore, we point osmotic stress, respectively), with brightfield and fluorescence out the relevance of dynamic aspects for the bistability of the Lac microscopy. We have monitored the rate of nuclear localization of system, which has recently been questioned in the natural context Mig1-GFP, Yap1-GFP, Hog1-GFP and Msn2-GFP in single cells, of E. coli by Dreisigmeyer et al. [2]. showing heterogeneity of the cell response that would be lost with [1] J. Yu et al., Science 311, 1600 (2006). [2] D. W. Dreisigmeyer conventional techniques. et al., eprint q-bio/0802.1223v1. Conclusions: In our experimental approach the environmental control is precise and reversible, allowing data on a single cell DS3-1-11 level to be acquired with high spatial and temporal resolution. Subsequent image analysis allows us to extract information useful Time dependent gene expression dynamics in Escherichia to elucidate how various signalling pathways work, including cell- coli batch culture to-cell variation, by measuring the protein kinetics. Dong, Guangqiang1; Bagh, Sangram2; Mandal, Mahuya2; David, McMillen2 DS3-1-13 1University of Toronto Mississauga, Department of Chemical and Physical Science, Mississauga, Canada; 2University of Toronto Informative single cell expression patterns in early Mississauga, Mississauga, Canada mammalian development Huss, Mikael; Guo, Guoji; Luo, Wenlong; Yeo, Zhenxuan; Clarke, With the advent of systems and synthetic biology, quantitative Neil; Robson, Paul understanding of gene expression and gene expression noise Genome Institue of Singapore, Singapore, Singapore gains high attention. In practice, to quantify and compare the gene expression level and variability of synthetic circuits in Objective: We are investigating gene regulatory networks in Escherichia coli, bacterial cells were transferred from overnight early development using mouse embryos and human embryonic culture into new fresh medium, grown till the early or mid- stem cells (hESC) as model systems. The focus is on the first exponential growth phage assuming a steady state expression differentiation events in the developing blastocyst. Real-time level and analyzed. However, no systematic studies were done qPCR (in the mouse) and genome-wide expression BeadChips (in on time dependent dynamics of gene expression and gene hESC) have been used to generate unique data on mRNA levels expression noise. Therefore how the gene expression level at various stages of early development. The objective of the work

176 ICSB 2008 presented here is to infer predictive gene expression patterns DS3-1-15 from the mouse embryo data. Results: In one project, 800 transcription factors (TFs) in the A trial to assess susceptibility of biomphalaria alexandrina preimplantation mouse embryo were screened to determine to schistosoma mansoni by Protein electrophoresis, those most likely to be involved in cell lineage specification. This hepatic enzymes and random amplification polymorphic screen selected 48 TFs, which were subsequently assayed by DNA (RAPD) real-time qPCR in single cells of a 3.5 day-old embryo. Principal Gawish, Fathyia1; Abu El Einin, Hanaa2; EL Bardicy, Samia2; component analysis on this data set on gene expression in Tadros, Menerva2 individual cells (rather than a pooled population) suggests that 1Theodor Bilharz Research Institute, Environmental Research and the expression patterns of the TFs are related to cell lineage Medical Malacology, Cairo, Egypt; 2Theodor Bilharz Research (epiblast, trophectoderm or primitive endoderm) even at this early Institute, Medical Malacology, Cairo, Egypt stage, when epiblast and primitive endoderm cells cannot be morphologically distinguished from each other. The first principal Objective: The identification of susceptible from refractory component, which seems to separate the trophectoderm lineage Biomphalaria alexandrina snails is important in the transmission from the others, has high loadings for known trophectoderm- studies and control programs of Schistosoma mansoni. associated genes like Cdx2. The second principal component has Results: In this study, three methods were used to asset the two high loadings for genes such as Gata4 that characterize primitive types of snails from each other. The first one was by using protein endoderm. electrophoresis in which many different bands were appearing Conclusions Our unusual single-cell data set allows us to in both susceptible and refractory snails, with one shard band get a sense of the amount of variation in TF expression levels between them (68.41KDa). The second method is the assay for between embryonal cells, and to investigate whether some of it is hepatic enzyme activities of aspartate aminotransferase (AST) and systematically related to cell lineage specification. Thus, we can alanine aminotransferase (ALT) as well as the Alkaline (ALKP) and derive composite expression profiles (instead of single marker acid phosphatase (AcP), In this method, a higher concentration genes) that characterize cells of different lineages and apply this of these enzyme were detected in tissue homogenate and to single blastomeres earlier in development to identify the initial hemolymph of susceptible snails than refractory snails. The third transcriptional events in early differentiation. method was by using Random amplification polymorphic DNA (RAPD), utilizing OPA-9 primer, approximately 1,900 bp while was DS3-1-14 characteristic for susceptible snails but not appeared in refractory one. Timing and dynamics of single cell gene expression in the Conclusions: in this study, susceptible B. alexandrina snails arabinose utilization system were identified by protein, enzyme and moleculer method, Posters Megerle, Judith1; Fritz, Georg2; Gerland, Ulrich2; Jung, Kirsten3; many moleculer investigation will be done on refractory snail to Dedicated Rädler, Joachim1 understand the genetics involved in this complex host/ parasite 1LMU München, Department für Physik und CeNS, München, relationship, and may be lead to an often discussed approach of Germany; 2Universität zu Köln, Institut für Theoretische Physik, introducing resistant B. alexandrina into the field as a means of Köln, Germany; 3LMU München, Department Biologie I, biological control for the parasite. München, Germany DS3-1-16 Objective: The arabinose utilization system of E.coli displays all- or-nothing gene expression at intermediate sugar concentrations ActivinβB, GDF11 and Follistatin participate in a feedback [1]. This stochastic effect divides an isogenic population into network that regulates genesis of neurons and glia by bacteria that express the system at a high level and bacteria that olfactory stem cells do not express the operon at all. We test whether the dynamics Gokoffski, Kimberly1; Wu, Hsiao-Huei2; Beites, Crestina L.3; of gene expression in single cells is compatible with the notion Lander, Arthur D.4; Calof, Anne L.3 that the stochastic nature of the basal expression of the arabinose 1University of California, Irvine, Dev/Cell Biology and CCBS, uptake proteins gives rise to the all-or-nothing behaviour [1]. Irvine, United States; 2Vanderbilt University School of Medicine, Results: Quantitative time-lapse fluorescence microscopy is Department of Biochemistry, Nashville, United States; 3University used to monitor the gene expression of individual cells following of California, Irvine, Department of Anatomy and Neurobiology, the addition of arabinose. Upon induction with a high arabinose Irvine, United States; 4University of California, Irvine, Center for concentration gene expression almost immediately starts in Complex Biological Systems, Irvine, United States all bacteria. In contrast, the onset is significantly delayed and varies from cell to cell at lower inducer concentrations. Using Objective: The olfactory epithelium (OE) is a sensory an analytical model, which is based on the assumption that neuroepithelium that avidly regenerates. Neurons are continuously arabinose uptake and GFP expression proceed in an independent generated to balance slow turnover and maintain a constant and consecutive manner we extract the time delay between number, but can also be generated rapidly in response to acute inducer addition and expression onset for each cell. This neuron loss. Rapid regeneration and maintenance of proper cell procedure is facilitated by prior analysis of the model parameters, numbers can be understood as distinct tasks that the OE has including the measurement of the GFP maturation time in evolved to perform. We seek to understand the control strategies single bacterial cells. The shape of the experimental time delay used to meet these performance objectives. distributions is in quantitative agreement with a stochastic model Results: We find that developmental neurogenesis in the OE is of the arabinose uptake process. Variations of the arabinose regulated by two feedback inhibitory factors that act at distinct concentration shift the delay to longer times, but do not change stages in the neuronal lineage: ActivinβB (ActβB), a secreted the shape of the distribution. protein produced by all OE neuronal cells, inhibits proliferation Conclusions: Our findings support the idea that the of stem and early progenitor cells. Growth and differentiation heterogeneous timing of gene induction is causally related to a factor 11 (GDF11), a protein related to ActβB, is secreted by broad distribution of uptake proteins at the time of sugar addition. late progenitor cells (INPs) and immature neurons, and inhibits The differences in timing might be beneficial for a bacterial proliferation of INPs. Activities of both ActβB and GDF11 are population by preventing costly synthesis of the arabinose system antagonized by follistatin (Fst), which is secreted by cells within in all cells when only modest amounts of arabinose are available. and underlying the OE. Absence of ActβB and GDF11 together [1] Siegele, D. A., and J. C. Hu, 1997, Proc. Natl. Acad. Sci. USA results in expansion of all cells of the neuronal lineage as well as 94:8168–8172 an increase in glial cells—a population not previously known to be controlled by either factor. Conclusions: Our findings support a model in which ActβB, GDF11 and Fst form a multistage feedback network that regulates genesis of neurons and glia by olfactory stem cells.

ICSB 2008 177 Modeling of this regulatory network suggests that these effectors DS3-1-19 contribute to an integrated feedback system in which the actions of ActβB are more closely connected to the achievement of Suboptimal metabolic behavior of Bacillus subtilis is steady state robustness (accuracy), while those of GDF11 are caused by developmental processes more closely associated with speed. Experiments are being Revelles, Olga; Heinemann, Matthias; Sauer, Uwe carried out to test these predictions in adult OE. Understanding ETH, Institute of Molecular System Biology, Zurich, Switzerland regulation of regeneration provides insight into the basic Although, population heterogeneity has been reported for biology of stem cells and their potential use in therapies for stressed isogenic Bacillus subtilis cultures, little is known about neurodegenerative diseases. its presence during the exponential phase of growth and its possible impact on the global physiology of the culture. A DS3-1-17 recent study reported that metabolism of B. subtilis becomes more optimal, e.g. high growth rates and yields, when post- Differences in the speed of transcription elongation exponential regulators that control the outcome of sporulation, between cells results in extrinsic noise in gene expression competence or/and motility are deleted. Here we address the Neves, Ricardo1; Andreu, Lorena2; Reittie, Joyce2; Gupta, Raj2; question whether these developmental regulators are involved in a Iborra, Francisco2 molecular mechanism that maintain metabolism in a sub-optimal 1University of Oxford, Molecular Haematology, Oxford, United state or whether it is due to the presence of subpopulations. Kingdom; 2University of Oxford, Molecular Haematology, Oxford, Quantitative analysis of a set of structural mutants that are United Kingdom defective in the above post-exponential processes indicated that only mutants with a motility minus phenotype grow more Identical cells in the same environment show significant optimal relative to the wild type. By using gfp fusion to hag, a phenotypical variability. Here we studied the source of global specific promoter of flagella formation, we quantified promoter extrinsic variability in gene transcription (global extrinsic noise) in expression of motility. We showed that the cascade that leads to mammalian cells. Previous studies have used protein reporters the formation of flagella is active during the exponential phase. to investigate noise in gene expression, obscuring the noise in These data indicate that sub-optimality in B. subtilis is due to the transcription. Here, we employed three more direct approaches to flagella formation cascade active during the exponential phase Dedicated study noise in transcription, involving the analysis of the dynamic of growing. Furthermore, quantitative single cell analysis by flow Posters behaviour of RNA polymerase II tagged with GFP (GFP-RNA pol cytometry indicates that only a subpopulation is motile. Currently II), the exchange of histone H2B tagged with GFP, and the use of we are testing whether subpopulation heterogeneity is causally Bromo-Uridine (BrU) as a tracer for transcription. related to the optimality phenotype. We demonstrate that individual cells in a homogeneous population transcribe at different speeds which are maintained Dedicated session 3-2: Synthetic biology over the time (hours). This study demonstrates that differences in transcription elongation are the primary source of extrinsic noise in mammals. DS3-2-10

DS3-1-18 CompuBioTic, A synthetic biology approach for diagnosis Rialle, Stephanie; Molina, Franck; Thierry, Alain Accuracy and robustness in procaryotic gene expression CNRS, Montpellier, France Tabaka, Marcin Institute of Physical Chemistry PAS, Department of Soft Objective: Synthetic biology aims at using biological parts in Condensed Matter, Warsaw, Poland order to construct new programmed biological systems. The ambition of our project is to establish a methodology to design Understanding biological module involves recognition of its and construct synthetic biological systems made of protein structure and dynamics of its principal components. We will elementary bricks and encapsulated within a non-living vesicle. present an analysis of dynamics of the repression module within Results: Within this perspective, we are constructing a the regulation of prokaryotic gene expression. We study by catalogue of standardised and reusable protein parts for computer simulation the effect of mode of corepressor binding synthetic biology. Each element from the catalogue has a well- to repressor and repressors to operator on accuracy of genetic characterized (formally and experimentally) role compatible switches and their stability with regard to corepressor fluctuations. with synthetic biological system engineering. An example is the It appears that a high-quality, accurate and effective module of role of “conditional sensor”, which refers to molecules able to repression is achieved if corepressor can attach/detach repressor- detect a biological signal and to give a conditional response operator complex and operator contains multiple binding sites for to this signal. Each element is described with a biological repressors. Multiple sites are responsible for stability of genetic processes formalisation language, BioΨ [1, 2]. It allows a switches in case of high corepressor concentration. If a decay formal and standardized description of the biological functions rate of repressor(s)-operator complex is low the switch is stable and processes at different levels of abstraction, and will allow under fluctuations of corepressor concentration even for small automatic identification of biological components of interest in a number of repressors in the cell. The accuracy of turning on of biological database. It is also appropriate to dynamical (temporal gene expression due to sudden uptake of corepressor is highly and spatial) modelling and simulations. For this work, we used the dependent on the rates of repressor(s) detachment form operator Hsim software, a stochastic cellular automaton, and then studied since the transcription and translation are inherently accurate. The the behaviour of the modelled system. We also characterized corepressor can detach from the repressor-operator complex and experimentally the parts of the catalogue and give them a score of destabilize it. Therefore, the whole process of gene expression is “reusability” for synthetic biology. accurate. Conclusions: The catalogue counts fifteen main entries classified The proposed model for regulation of gene expression by with respect to the scale of the system. A theoretical formal repression is widely encountered in the cell of E. coli. For example description of all the parts is available and some of them have repressors TrpR, MetJ, ArgR or PurR regulate genes expression been assay and characterized experimentally. For instance, in described manner. In order to confirm our predictions we will we have encapsulated molecules implementing the “revealing” present detailed description of the repression of trp operon in E. role entry in liposomes, and characterized it experimentally. We coli. designed and model a theoretical system dedicated to early medical diagnosis task. [1] Mazière, P., Granier, C., Molina, F., (2004) J. Mol. Biol., 339:77- 88 [2] Mazière, P., Parisey, N., Aimar, M., Molina F., (2007), J. Biosciences, 32(1)145-15

178 ICSB 2008 DS3-2-11 efficiency, degradation rate of proteins, UV induction etc. Experimentally we have constructed those gene circuits and Learning robustness from gene regulatory networks implemented in the lysogenised (K-12 wild type) cell. Our initial Dubrova, Elena results showed that those circuits were able to detect and Royal Institute of Technology, Electronics, Computer and Software prevent the bacterial lysis as in-cell disease prevention device. Systems, Kista, Sweden The logic level of the circuits was measured by the fluorescence of the polycistronically attached reporter protein, enhanced green Objectives: Our long-term ultimate goal is to be able to design fluorescence protein by flowcytometry. The relative amount of and manufacture engineered biological systems that perform death and survival of the bacteria were measured by optical computation in a way electronic systems do today. A fundamental density at 600nm. challenge here is the synthetic creation of a robust system capable of operating under changing environments and noisy DS3-2-14 input, and yet exhibit the desired behavior under constraints such as power consumption, area, and performance. Natural Heterologous production of nonribosomal peptides in biological systems are able to handle these challenges with an Saccharomyces cerevisiae elegance and efficiency. Our short-term objective is to learn more Chen, Xiao about how gene regulatory networks of living cells perform robust Technical University of Denmark, Systems Biology, Lyngby, computation and to attempt to mimic it. Denmark Results: Previous work have shown that the robustness of gene regulatory network is due to the intrinsic stability of their Objective: Bioactive peptides, which are usually synthesized attractors’ landscape [1]. We propose a computational scheme nonribosomally, are a large and diverse group of natural products which exploits the stability of attractors. In this scheme, the states with great significance to human pharmaceuticals. Due to the of a network represent variables of the computed function, and scarcity of their production in nature and the complexity of these attractors represent function’s values. We test the robustness of molecules, purification from biological material or chemical the presented computational scheme on the example of a gene synthesis is inherently difficult. Herein, we attempted to use regulatory network model of Arabidopsis thaliana [2] and show an alternative method to produce non-ribosomal peptides in a that it can tolerate 91.58% single-point mutations in the outputs microbial host, Saccharomyces cerevisiae. For this purpose, we of the defining tables of gene functions. chose nonribosomal tripeptide L-amino-adipate-L-cysteine-D- Conclusions: Traditionally, robustness of a system is achieved by valine (ACV), the precursor of penicillins and cephalosporins, as adding redundant components. We show that there are cheaper our target in this study. alternatives. In the computational scheme which we introduce, Results: The Penicillium chrysogenum gene pcbAB encoding Posters the robustness is due to the non-uniqueness of paths leading ACV synthetase (ACVS) was cloned and expressed in S. Dedicated to an attractor. A fault may change a path, but the destination cerevisiae, together with Aspergillus nidulans gene npgA remains the same with a high probability. Therefore, if attractors encoding phosphopantetheinyl transferase (PPTase), which is are stable, the network is able of sustain the majority of faults. required to convert the expressed ACVS to its active form. The Bibliography: production of ACV was detected by LC-MS in yeast and was [1] Robustness and evolvability in gene regulatory networks, M. subsequently enhanced by decreasing cultivation temperature Aldana, E. Balleza, S. Kauffman, O. Resendis, Journal of Theor. from 30 degree to either 20 or 15 degree. Codon optimization of Biol., vol. 245, 2007, pp. 433-448. the ACVS encoding gene, pcbAB, also improved ACV production [2] From Genes to Flower Patterns and Evolution: Dynamic in some extent in yeast. Models of Gene Regulatory Networks, A. Chaos, M. Aldana, C. Conclusions: For the first time, nonribosomal tripeptide ACV Espinosa-Soto, B. Garcia Ponce de Leon, A. G.Arroyo, E. R. was successfully produced in S. cerevisiae after introducing Alvarez-Buylla, Journal of Plant Growth Regulation, vol. 25, no. 4, the fungal biosynthetic genes into yeast. Our data showed that 2006, pp. 278-289. both low-temperature cultivation and codon optimization of the gene encoding ACV synthetase had enhanced the heterologous DS3-2-13 production of ACV in yeast. The work here will not only accelerate research on production of In-cell disease prevention device desired nonribosomal peptides in yeast, but also facilitate a better Bagh, Sangram; McMillen, David understanding of the synthetic biology of S. cerevisiae. University of Toronto, Chemistry, Mississauga, Canada DS3-2-16 Objective: One of the distant dreams of synthetic biologists is to develop a “programmed gene circuit” which could act as an The era of genome wide analysis and comparative in-cell disease detection and prevention device. We showed that transcriptomics in Aspergillus this could be done in bacterial cell using an electrical engineering Salazar, Margarita1; Nielsen, Jens2 approach. 1Technical University of Denmark, Center for Microbial Results: Bacteriophage lambda is a virus, which attacks Biotechnology, Biosys, Kgs. Lyngby, Denmark; 2Chalmers the bacteria Escherichia coli and establishes a biochemical University of Technology, Systems Biology, Gothenburg, Sweden lysis-lysogeny switch. In lysogeny, the virus writes itself into the bacterial genome, and divides along with the host, doing Objective: With the publication of the genome sequences of essentially no harm. Certain events like UV induction can trigger the three aspergilli, Aspergillus nidulans, A. oryzae and A.niger, a switch into the lysis pathway; where the virus replicates very genome wide systems biology studies are now possible. New fast and then breaks the cell wall and kills the host. From an high-density arrays from Affymetrix were designed for these three electrical engineering perspective, this switch can be viewed as aspergilli in order to study transcriptional responses. R-S latch circuit. We have designed several artificial gene circuits, Results: Combining the genome scale models with which can detect, couple and control the function of lysis- transcriptional data in the context of metabolism, overall lysogeny switch. Those circuits stay dormant in no-harm lysogeny metabolic responses were addressed and eventually moved state but are able to detect the relative change in output levels towards refining the existing models for growth, organic acid (concentration of protein CI and Cro) of the lysis-lysogeny switch. and protein production. Genes having homologues in both Upon detection, the circuit is turned on to prevent the lysis by species were identified using a blastp-based comparison. Not expressing CI protein to block the expression of genes necessary surprisingly, A. niger and A. oryzae were more closely related to to access the lysis pathway. each other than the other Aspergillus, and at least 5561 ORFs Conclusion: We generate simple mathematical models to predict showed to be conserved. A comparative transcriptomics study the behavior of those circuits as a function of different biochemical was conducted, based in A. nidulans, A. niger and A. oryzae. parameters such as strength of ribosome binding site, repression With the establishment of a set of fermentation experiments using

ICSB 2008 179 glucose or xylose as a carbon source, it allowed the identification Dedicated session 3-3: Software tools of 23 conserved genes that were differentially regulated in all three species and which seem to be a conserved response across the DS3-3-09 Aspergillus genus. The homologues were regulated in the same direction. The function of the genes was unknown in A. oryzae Arena on the systems biology toolbox 2 for MATLAB and A. nidulans, but due to the improved annotation of A. niger, Schmidt, Henning; Frey, Simone it was possible to assign the function of the other two species University of Rostock, Rostock, Germany based on that of A. niger and the conserved sequences and transcription responses. The xylanolytic transcriptional activator The Systems Biology Toolbox 2 (SBTOOLBOX2) for MATLAB XlnR, previously only described in A. niger and A. oryzae, had a offers systems biologists a powerful, open, and user extensible homologue in A. nidulans, and was significantly induced on xylose environment, in which to build models of biological systems. as well. This result confirmed that the XlnR regulation is present in Experiments can be performed on models, just like in real lab A. nidulans and functioning in a similar manner to that reported for life but in silico. The representation of models, experiments, A. niger and A. oryzae. Moreover, this experiment was also good and measurement data is intuitive and easy to use. The toolbox for validating non specific hybridization of fungal cRNA for each of features a wide variety of specialized analysis tools that can the species in the probes that did not correspond to that specie. directly be applied in the modeling process. MATLAB adds to that Conclusions: Comparative transcriptomics for these aspergilli by a large number of inbuilt functions and a high level scripting is now a tool and extensively used to investigate different cellular language, allowing the user to quickly and efficiently add new mechanisms in their metabolic and regulatory levels. functionality. The SBPD extension package for the SBTOOLBOX2 adds high- DS3-2-19 speed simulations, combination of models, experiments, and measurement data in so called projects. Functions are available Inferring closed-loop responses from open-loop that support the complete model building process (modeling, characteristics for a family of synthetic transcriptional simulation, identifiability analysis, model reduction, parameter feedback systems estimation (multiple experiment and multiple measurement fitting), Rai, Navneet1; Dhabolkar, Sugat2; Ramkumar, Krishna3; validation, etc.). The projects are a powerful construct that allows Dedicated Sreenivasan, Varun4; Venkatesh, K.V.5; Thattai, Mukund2 keeping a perfect overview over your modeling tasks at any time. Posters 1Indian Institute of Technology Bombay, School of Biosciences Graphical user interfaces support the workflow. and Bioengineering, Mumbai, India; 2National Centre of Biological Over the last 3 years the Systems Biology Toolbox for MATLAB Sciences, Bangalore, India; 3Indian Institute of Technology has become widely used for modeling and analysis of biochemical Bombay, Chemistry, Mumbai, India; 4St. Xavier’s, Mumbai, India; systems in the area of systems biology. Since January 2008 5Indian Institute of Technology Bombay, Chemical Engineering, the considerably enhanced SBTOOLBOX2 is available that, Mumbai, India despite its recent publication, already has shown to be a very popular download amongst modelers in both academia and the Synthetic biology seeks to build biological systems from the pharmaceutical industry. bottom up: to design and construct complex systems from The arena on ICSB-2008 will allow participants of the conference simple parts. Essential to the design process is the ability to that do not have the possibility to participate in the SBTOOLBOX2 predict the response of a system by measuring the behavior of tutorial to interact with the developers of the toolbox, to provide its parts, even as the complexity of the system is scaled up. Our feedback and to obtain information. Real life example modeling project is designed so that, by measuring the behavior of relatively projects can be experienced in live demonstrations. simple systems called ‘open loops’, we are able to predict the The Systems Biology Toolbox 2 is free software and can be response of more complex systems called ‘closed loops’. If these downloaded at: www.sbtoolbox2.org predictions match the observed responses, it would rigorously Contact demonstrate that we can indeed go from parts to systems. Henning Schmidt, PhD, University of Rostock, Germany, For this proof-of-principle demonstration, we chose to use [email protected] transcriptional regulatory networks as our model systems, with Simone Frey, PhD Student, University of Rostock, Germany, their steady-state responses being the behavior of interest. [email protected] Recent theoretical work [Angeli D, Ferrell JE Jr. and Sontag ED (2004), PNAS 101, 1822] has shown that the steady-state DS3-3-11 response of a transcriptional regulatory system involving feedback (closed loop) can be predicted by characterizing its response GraphCrunch: A tool for large network analyses when feedback is cut (open loop). This technique scales up: it Przulj, Natasa; Milenkovic, Tijana; Lai, Jason can be applied to arbitrarily complex systems, as long as they University of California, Irvine, Computer Science, Irvine, United satisfy certain simple conditions. We set out to test this method States experimentally. We used three transcriptional regulators (LacI, TetR, and the Objective: The recent explosion in biological network data has LuxI/LuxR module) and two fluorescent reporters (CFP and created the need for improved tools for large network analyses. YFP) to implement our open and closed loops in the bacterium In addition to well established global network properties, small Escherichia coli. To test the generality of the method, we induced subgraphs, called graphlets, have been used to develop designed our open-loop system with two inputs and one output, “network signatures” that summarize network topologies. Based so it could be ‘closed’ in two different ways: with either LuxR or on these network signatures, two new highly sensitive measures LuxI in feedback. We used fluorescence microscopy to measure of network local structural similarities were designed: the relative the open-loop characteristic, and applied the method of Angeli et graphlet frequency distance (RGF-distance) and the graphlet al. to predict the closed-loop response. Interestingly, the closed degree distribution agreement (GDD-agreement). loop with LuxR in feedback is predicted to be monostable, while Finding adequate null-models for biological networks is important that with LuxI in feedback is predicted to be bistable. Finally, we in many research domains. Network properties are used to measured the actual closed-loop response of the LuxR feedback assess the fit of network models to the data. Various network system, and found that it was consistent with our predictions. models have been proposed. To date, there does not exist a This represents the first experimental validation of this powerful software tool that measures the above mentioned local network bottom-up design principle. properties. Moreover, none of the existing tools compare real- world networks against a series of network models with respect to these local as well as a multitude of global network properties. Results: Thus, we introduce GraphCrunch, a software tool that finds well-fitting network models by comparing large real-world networks against random graph models according to various

180 ICSB 2008 network structural similarity measures. It has unique capabilities gene expression responses are addressed as case studies. of finding computationally expensive RGF-distance and GDD- agreement measures. In addition, it computes several standard DS3-3-15 global network measures and thus supports the largest variety of network measures thus far. Also, it is the first software tool that Automated live cell image segmentation for on-line high- compares real-world networks against a series of network models content screening and that has built-in parallel computing capabilities allowing for Paul, Perrine1; Kalamatianos, Dimitris1; Duessmann, Heiko2; a user specified list of machines on which to perform compute Huber, Heinrich2 intensive searches for local network properties. Furthermore, 1National University of Ireland, Hamilton Institute, Maynooth, GraphCrunch is easily extendible to include additional network Kildare, Ireland; 2RCSI, Department of Physiology and Medical measures and models. Physics, York House, York Street, Dublin 2, Ireland Conclusions: We present GraphCrunch as a comprehensive, parallelizable, and easily extendible software tool for analyzing Objective: Our project focuses on the development of a novel and modeling large biological networks. The software is open- monitoring system for live cell confocal microscopy that is source and freely available at http://www.ics.uci.edu/~bio-nets/ applied to study the biochemical signal transduction during graphcrunch/. programmed cell death. We thereby exploit image analysis techniques together with biological markers to gain information DS3-3-12 about cellular processes from image time series. To do so, cells are automatically segmented and tracked during the whole PANTHER pathway applet – an SBGN compatible pathway experiment to obtain time-series information per cell. visualizing tool based on CellDesigner 4.0 Results: Because minimal user input and fast processing times Mi, Huaiyu1; Muruganujan, Anushya1; Masuoka, Yukiko2; are required for real time processing, the user only interacts Funahashi, Akira3; Kitano, Hiroaki2; Thomas, Paul1 with the system during the set up process setting the range 1SRI International, Evolutionary Systems Biology Group, Menlo of cells diameter and picking the algorithm with the best Park, United States; 2The Systems Biology Institute, Tokyo, segmentation results among the following options: local maxima Japan; 3Keio University, 3Department of Fundamental Science seeded watershed [1], modified constrained erosion dilation and Technology, Yokohama, Japan [1], and circular seeded watershed [2]. In all cases, images are automatically preprocessed to remove noise and enhance The PANTHER (Protein ANalysis THrough Evolutionary contrast. Relationships) Classification System was designed to model As a key feature, the segmentation algorithms allow the evolutionary sequence-function relationships on a large scale. Its concurrent evaluation of several image channels that are Posters core is a library of protein families that have been subdivided into generated by simultaneously measuring different cellular probes Dedicated functionally related subfamilies, using human expertise. These like mitochondrial stain, cytosolic fluorescent protein and subfamilies model the divergence of specific functions within Differential Interference Contrast (DIC). Results were obtained with protein families, allowing more accurate association with function, HeLa cervical cancer and DU colon cancer cells using a confocal as well as inference of amino acids important for functional microscope (Zeiss LSM510) and a combination of DIC with specificity. Hidden Markov models (HMMs) are built for each channels detecting the fluorescent protein and the mitochondrial family and subfamily for classifying additional protein sequences. stain. PANTHER pathway (http://www.pantherdb.org/pathway/) is Conclusions: Despite the high complexity of segmentation due one of the modules of the PANTHER System. It contains a large to aggregated cells and noise, segmentation results satisfying collection of expert curated metabolic and signaling pathways. the live imaging users were achieved in most cases with a All pathways were created with CellDesigner, an SBGN (Systems maximum processing time of 10 seconds. The segmentation tool Biology Graphical Notation) compatible pathway network editing was implemented in a stand alone demo version, but in a later software. stage, this system will be extended to perform a fully automated, PANTHER Pathway Applet is a java applet that displays situation dependent, set-up of the microscope that aims to PANTHER pathways on the PANTHER website. The applet speed-up and facilitate experimental analysis by reducing manual displays two views of pathways: The standard view based on intervention. SBGN Process Diagram level 1 notation and the light view References: based on SBGN Activity Flow notation. The PANTHER pathway [1] PS Umesh Adiga & BB Chaudhuri. Comput Methods applet code was developed as an extension to the CellDesigner Programs Biomed, 2000, 61, 23-47 4.0. There are three main components to the applet: Subset [2] A Thomas, T Davies & A Luxmoore. Anal Quant Cytol Histol, of libSBML functionality written in Java to parse and create 1992, 14, 347-53 pathway objects, extensions to the CellDesigner application and functionality to generate and render the light view. Furthermore, DS3-3-16 there is functionality to select the upstream and downstream path for any given component in a pathway diagram. Structured and modular modeling with ProMoT Steinmetz, Katrin; Mirschel, Sebastian; Rempel, Michael; Martin, DS3-3-13 Ginkel; Gilles, Ernst Dieter Max Planck Institute for Dynamics of Complex Technical Systems, Polar Mapper: Computational tool for integrated Systems Biology Group, Magdeburg, Germany visualization of protein interaction networks and mRNA expression data Objective: In recent years Systems Biology has become a Valente, André X. C. N.1; Gonçalves, Joana P.2; Grãos, Mário3 massive information science, driven by data that is produced by 1Biocant and University of Coimbra, Unidade de Sistemas powerful high throughput technologies. Based on these data, Biológicos, Cantanhede, Portugal; 2Biocant, IST - Technical models are growing fast in terms of size, number and complexity. University of Lisbon and INESC-ID, Cantanhede, Portugal; Thus, modular modeling concepts become more and more 3Biocant, Unidade de Biologia Celular, Cantanhede, Portugal attractive. ProMoT, our open-source modeling tool is aimed to support Objective: Polar Mapper is a computational application ideally structured and modular modeling of both quantitative models suited for exposing the architecture of protein interaction described by differential-algebraic equations and qualitative networks. It further facilitates the natural system-level analysis of models based on a logical model formalism. mRNA expression data in the context of the underlying protein Results: Besides already established features such as the interaction network. modular modeling which can be done in an object-oriented Results: Preliminary analysis of a human protein interaction fashion based on extensible modeling libraries and reusable network and comparison of yeast oxidative stress and heat shock modules and the exchange of models with other tools using

ICSB 2008 181 SBML, new functionality has been added. DS3-3-18 A central point is to improve the modeling workflow by using (automatic) modularization tools accompanied by advanced Usability improvement for visual editing in the modeling visualization methods and usability concepts like validation of tool ProMoT the module structure and shortcuts for editing. Newly added Steinmetz, Katrin; Ginkel, Martin; Samaga, Regina; Gilles, Ernst functionality comprises a tool which has been developed to Dieter automatically structure SBML models into separate modules. Max Planck Institute for Dynamics of Complex Technical Systems, This can be used as a preprocessing step before the import of Systems Biology Group, Magdeburg, Germany a model into ProMoT. Furthermore, a modularization method is provided, which can be used to conveniently relocate selected Objective: Usability of Software becomes increasingly important components in a new, separate module to hide complexity or in interdisciplinary research areas such as Systems Biology reuse it in a different context. In addition to the SBML support, because the work flow often demands the user to try and ProMoT offers further import and export facilities such as the use different tools. Therefore, a tool should be intuitive and model export to different simulation and analysis tools and the the learning curve should be low. ProMoT is a modeling tool export of the visual model representation to a variety of graphics for signaling, regulatory and metabolic networks in Systems and image formats. Biology. Modeling can be done in an object-oriented fashion Conclusions The presented new features are aimed to based on object-oriented extensible libraries. The models can conveniently handle highly structured and modular models be exchanged by the international standard language SBML. by modularization tools and advanced editing support and ProMoT consists of three main components, the Browser, the to intuitively illustrate large-scale and complex models using Visual Editor and the Visual Explorer. In former times ProMoT advanced visualization methods. Further information about used to be a very generic tool and could also be used for ProMoT and the download link can be found at modeling in the field of chemical engineering. Now it comprises http://www.mpi-magdeburg.mpg.de/projects/promot. rather specific functionalities for models in Systems Biology, in particular for a new modeling formalism of logical models, which DS3-3-17 has recently been introduced. Results: Because of complex modeling libraries, shortcuts have Dedicated Visual support for structural and functional analysis of been introduced and implemented to support the modeler in first Posters complex signaling networks in ProMoT modeling steps and also to speed up advanced modeling. In the Mirschel, Sebastian1; Saez-Rodriguez, Julio2; Ginkel, Martin1; shortcuts exemplary for activation and inhibition elements are Gilles, Ernst Dieter1 inserted automatically. Furthermore, an online-checker-system 1Max Planck Institute, Systems Biology Group, Magdeburg, has been implemented to show modeling errors, which would Germany; 2Harvard Medical School, Department of Systems maybe not be recognized until the export or analysis of the model. Biology, Boston, United States The modeling errors are namely falsely used library elements, wrong connected library elements or incomplete network parts. Objective: In recent years signaling aspects of biological systems The modeler gets alerted by different colored boxes emphasizing become more and more popular. Based on massive data the elements and can obtain a more specific textual description of sets, signaling networks are fast growing in terms of size and the error. complexity. Thus, the structural and functional analysis of large Conclusions: A lot of new features improve the usability of signaling networks using a logical (Boolean) model formalism is ProMoT for modeling in Systems Biology. More information about a valuable approach. In contrast to more detailed, quantitative ProMoT and the software can be obtained on our homepage descriptions, the simplicity of this approach allows to handle (http://www.mpi-magdeburg.mpg.de/projects/promot) considerable systems and couple models to large sets of data. However, the interpretation of such networks and their analysis DS3-3-19 may not be trivial as they rely on the properties of complex networks. Therefore, adequate visual presentations would be of CellDesigner4.0: A process diagram editor for gene- great value. regulatory and biochemical networks Results: We present an approach aimed to provide intuitive Funahashi, Akira1; Jouraku, Akiya2; Matsuoka, Yukiko3; Kikuchi, and flexible visual representations of analysis data. The results Norihiro4; Kitano, Hiroaki3 are directly mapped to and visually encoded in the analyzed 1Keio University, Dept. of Biosciences and Informatics, Yokohama, network. For example, it is feasible to visually group identical or Japan; 2Keio University, Yokohama, Japan; 3The Systems Biology similar data (e.g. to depict possible correlations between proteins Institute, Tokyo, Japan; 4Mitsui Knowledge Industry Co., Ltd., or functional groups), to emphazise non-trivial or unexpected Tokyo, Japan data (e.g. activation of a certain protein that was not expected) and to de-emphazise data that is not relevant in the context of Objective: Identification of the logic and dynamics of gene- a specific analysis (e.g. hide proteins that are not involved in a regulatory and biochemical networks by synergistic integration certain pathway). Another interesting feature is the simultaneous of theory, computational modeling and experiments is a major presentation of heterogeneous analysis results within a single challenge of systems biology. From the view of computational illustration by encoding them using different visual properties (e.g., modeling, a model is used to understand the dynamics of for a certain protein, the initial value and the value obtained from biological phenomena, which consists of molecules and reactions the analysis are mapped to the node color and the node label, that represents gene regulatory and biochemical network. By respectively). using this model, it would be possible to simulate the dynamics Conclusions: The presented approach supports the analysis of of the model and compare the simulation results with their complex signaling phenomena and the integration of interrelated experiments and even more, it would be possible to tune the analysis results using visual methods. These methods are part of parameter in the model to fit with the experimental results. a visual environment implemented in the modeling tool ProMoT, Development of software infrastructure to support this workflow is where models can also be set up and exported for analysis (Saez- essential for systems biology research. Rodriguez, Mirschel et al, BMC Bioinf, 7:506, 2006). ProMoT is Results: We have developed a modeling / simulating tool called freely available for download at http://www.mpi-magdeburg.mpg. CellDesigner which primarily has capabilities to visualize, model de/de/projects/promot. and simulate gene-regulatory and biochemical networks. Two major characteristics embedded in CellDesigner boost its usability to create / import / export models; that is, 1) Consistent and comprehensive graphical representation (SBGN: Systems Biology Graphical Notation) of network models, and 2) SBML (Systems Biology Markup Language) as a model-describing basis, which functions as inter-tool media to import / export models.

182 ICSB 2008 Simulation capabilities are supported by integration with SBW- introduces a freely downloadable, software package, SBML-SAT, enabled software packages and SBML ODE Solver. Users can which implements algorithms for simulation, steady state analysis, also browse and modify existing models by referring to existing robustness analysis and local and global sensitivity analysis for databases directly through CellDesigner. SBML models. Conclusions: We have developed CellDesigner, a process Results: This software tool extends current capabilities through diagram editor for gene-regulatory and biochemical its execution of global sensitivity analyses using multi-parametric networks based on standardized technologies and with wide sensitivity analysis, partial rank correlation coefficient, SOBOL’s transportability to other SBML-compliant applications and method, and weighted average of local sensitivity analyses in SBW-enabled modules. CellDesigner also aims to support the addition to its ability to handle systems with discontinuous events standard graphical notation. Since the standardization process is and intuitive graphical user interface. still underway, our technologies are still changing and evolving. Conclusions: This work provides the community of systems CellDesigner runs on multiple platforms such as Windows, Linux, biologists a new tool for the analysis of their SBML models of and Mac OS X, and is freely available from our website at http:// biochemical and cellular processes. celldesigner.org. References: Zi et al. SBML-SAT: A Systems Biology Markup Language (SBML) based Sensitivity Analysis Toolbox. 2008, DS3-3-20 Submitted Zi et al. In silico identification of the key components and steps Hubba-Hubba: Hub objects analyzer - A framework of in IFN-gamma induced JAK-STAT signaling pathway. FEBS Lett interactome hubs identification for network biology 2005, 579(5):1101-1108. Chen, Shu-Hwa1; Lin, Chung-Yen1; Chia-Hao, Chin1; Hsin-Hung, Zheng et al. Comparative study of parameter sensitivity analyses Wu1; Chin-Wen, Ho2; Ming-Tat, Ko1 of the TCR-activated Erk-MAPK signalling pathway. IEE Proc Syst 1Institute of Information Science, Academia Sinica, Taipei, Taiwan; Biol 2006, 153(4):201-211. 2Department of Computer Science and Information Engineering, National Central University, Chung-Li, Taiwan DS3-3-22

One major task in the post-genome era is to reconstruct myBLAST : A BLAST web service for customized databases proteomic and genomic interacting networks using high and result analyzer throughput experiment data. To identify essential nodes/ hubs in Lin, Chung-Yen; Chen, Shu-Hwa these interactomes is a way to decipher the critical keys inside Institute of Information Science, Academia Sinica, Taipei, Taiwan biochemical pathways or complex networks. These essential nodes/hubs may serve as potential drug-targets for developing BLAST is a well-developed method for finding similarity among Posters novel therapy of human diseases, such as cancer and infectious nucleotide sequences or amino acid sequences. However, in Dedicated disease caused by emerging pathogens. most BLAST service, user can only blast to the pre-existing Hub Objects Analyzer (Hubba-Hubba) is a web-based service for database with penalty options. With this intuitive web-based exploring important nodes in an interactome network generated interface named as myBLAST, we try to provide a web paltform from specific small-scale or large-scale experimental methods for blast to customized database. In myBLAST, users can upload based on graph theory. Two characteristic analysis algorithms, their own sequence collection in FASTA format, specify the name Maximum Neighborhood Component (MNC), and Density of and the type of the database to create their own databases. Maximum Neighborhood Component (DMNC) are developed for After the database has been formatted, the user will receive a exploring and identifying hubs/ essential nodes from interactome notification mail from system, and can start the blast search on networks. Users can submit their own interaction data in PSI the customized databases in web interface. The blast search in (Proteomics Standards Initiative, version 2.5 and 1.0), tab format myBLAST can take single or multiple sequences at a time via and tab with weight values. After calculation completed, user a copy-paste procedure or an uploaded sequence file in fasta will get an email notification in minutes or hours depended on format. All the results will be presented in a tabular layout for the size of interactomes. The result will include a rank given by navigation and the raw blast result as well as *.csv file are ready a composite index, a manifest graph of network to show the for downloading. With the built-in blast parser, users can extract relationship amid these hubs, and links for retrieving output files. the result more easily by the rank of hits. All the information related This proposed method can be applied to discover some with submitted works will be sent by e-mail to notify users to unrecognized hubs from previous dataset. For example, most of catch the results. myBLAST performs stably and efficiently in the the Hubba high-ranked hubs (80% in top 10 hub list, and >70% implementation of linux, mysql, php and Java. The construction of in top 40 hub list) from the yeast protein interactome data (Y2H a customized database of 5000 nucelotide sequences (average experiment) are reported as essential proteins. Since the analysis length= 1kb) and 56000 sequences (average length= 1.2kb) will methods of Hubba-Hubba are based on topology, it can also be just take 1 second and 65 seconds, respectively. The searching used on other kinds of networks to explore the essential nodes, time for a blast search in a batch of 500 or 1000 nucleotide like networks in yeast, rat, mouse and human. The website of sequences (average length= 1kb) are 34 seconds or 83 seconds Hubba-Hubba is freely available at http://hub.iis.sinica.edu.tw/ to a database constituted by 5000 sequences (average length= Hubba. 1kb). The framework of myBLAST is well designed for users doing management with simple mouse clicks. Notably, myBLAST is DS3-3-21 freely accessible at http://mybioweb.nhri.org.tw/yourblast/ and has started service from 18 months ago. SBML-SAT: A Systems Biology Markup Language (SBML) based sensitivity analysis toolbox DS3-3-23 Zi, Zhike1; Zheng, Yanan2; Rundell, Ann3; Klipp, Edda1 1Max Planck Institute for Molecular Genetics, Computational Mariner: A web application framework for PySCeS Systems Biology, Berlin, Germany; 2Entelos Inc., Foster City, Olivier, Brett; Rohwer, Johann; Hofmeyr, Jan-Hendrik United States; 3Purdue University, Weldon School of Biomedical University of Stellenbosch, Biochemistry, Stellenbosch, South Engineering, West Lafayette, United States Africa

Objective: It has long been recognized that sensitivity Objective: Rapid advancements in systems biology have led to analysis plays a key role in modeling and analyzing cellular and a concomitant increase in the number of analysis techniques that biochemical processes. Systems Biology Markup Language are available for modelling cellular systems. The general adoption (SBML) has become a well-known platform for coding and of the Systems Biology Markup Language (SBML) as a standard sharing mathematical models of such processes. However, for model description [1] and the ubiquitous nature of the internet current SBML compatible software tools are limited in their ability has allowed modelling tools to exchange models and, potentially, to perform global sensitivity analyses of these models. This work share services with each other. In this contribution we describe

ICSB 2008 183 Mariner, a web applications extension to the Python Simulator for DS3-3-25 Cellular Systems (PySCeS) [2]. Results: The Mariner server exposes the main PySCeS methods From escherichia coli microarray raw data to pathways and (simulation, steady state, control analysis) as SOAP web services published abstracts using taverna and web services (using the Python “soaplib” library, http://trac.optio.webfactional. Maleki-Dizaji, Saeedeh1; Holcombe, Mike1; Fisher, Paul2; Rolfe, com). Additional features include: an extendible design, model Matt3 object persistence as well as user defined model upload in both 1The University of Sheffield, Computer Science, Sheffield, United PySCeS and SBML formats. Kingdom; 2The University of Manchester, Computer Science, The Mariner client consists of two components, a remote PySCeS Manchester, United Kingdom; 3The University of Sheffield, client which allows users to interact with a Mariner server in a Department of Molecular Biology and Biotechnology, Sheffield, way that emulates the PySCeS console model object and a new United Kingdom module which provides access to e.g. the “Draw Network” and “SBML Layout” web services provided by the Systems Biology Objective: Many microarray studies have analysed data in a Workbench [3]. user-intensive manner to identify regulons, pathways and relevant Conclusions: Mariner is a new extension to PySCeS which literature. A two-colour cDNA microarray dataset comprising provides users the ability to leverage the diverse capabilities a time-course experiment of Escherichia coli cells during an of other systems biology software from a familiar modelling aerobic to anaerobic environment is used to demonstrate a data- environment. Additionally, it provides the broader systems biology driven methodology that identifies known pathways from a set of access to PySCeS functionality. differentially expressed genes. These pathways are subsequently [1] Hucka, M. et al. The systems biology markup language used to obtain a corpus of published abstracts (from the PubMed (SBML): A medium for representation and exchange of database) relating to each biological pathway identified. biochemical network models. Bioinformatics 19, 524-531 (2003). Results: The workflow consists of three parts: Microarray Data [2] Olivier, B. G., Rohwer, J. M. & Hofmeyr, J.-H. S. Modelling Analysis; Pathways extraction; and PubMed abstract retrieval, cellular systems with PySCeS. Bioinformatics 21, 560-561 (2005). which is implemented systematically through the use of web [3] Hucka, M. et al. The ERATO Systems Biology Workbench: services and workflows. For the purpose of implementing this Enabling interaction and exchange between software tools for systematic pathway-driven approach, we have chosen to use the Dedicated computational biology. Pac. Symp. Biocomput. 450-461 (2002). Taverna workbench. The “Microarray Data Analysis” part provides Posters This work has been funded by the South African National data loading, normalisation and significance testing of microarray Bioinformatics Network data, providing a range of diagnostic plots of the microarray data, including histograms, box plots and principal component analysis DS3-3-24 plots using R and Bioconductor. The “Pathway extraction” part searches for differentially-expressed genes and cross-references SemanticSBML - A tool for annotating, checking and them to the KEGG database to obtain gene and pathway merging of biochemical models in SBML format descriptions. The “PubMed abstract retrieval” part takes the Liebermeister, Wolfram; Krause, Falko; Uhlendorf, Jannis; Klipp, pathway descriptions and searches the PubMed database to Edda identify up to 500 abstracts related to the chosen biological MPI for Molecular Genetics, Berlin, Germany pathway. Conclusions: The result of this research is a re-usable Objective: A major practical challenge in systems biology is methodology that directly processes raw microarray data files and to combine existing models of biochemical networks. Many that can ultimately identify published abstracts (from PubMed) that structural and dynamic cell models are available in the exchange are relevant to the genes and pathways found during analysis of format SBML (System Biology Markup Language) and in model the microarray data. This provides biomedical researchers with an databases like BioModels.net or JWS online. Computer-assisted integrated system for the analysis of Affymetrix/cDNA microarray merging of such models would considerably facilitate large-scale experiments, makes analyses more accessible to biologists/ model building. Completely automatic model merging is likely to medics rather than solely bioinformaticians, reduces training fail, though, even with simple real-world examples. Therefore, requirements and time, improves the productivity of bioinformatics a practical merging software should be interactive: it should be array support staff, reduces the number of errors associated with able to detect conflicts between models and assist researchers in manual data analysis and improves the reproducibility of methods. resolving them. Results: The program semanticSBML (http://sysbio.molgen. DS3-3-26 mpg.de/semanticsbml/) helps users to annotate, check, and combine models in SBML format. In addition, the software OCPID: Organ Centric Protein Interaction Database allows to build SBML models automatically from a list of KEGG Lee, Hyunju; Han, Beomsoo reaction identifiers and to visualise their network structure. It is Gwangju Institute of Science and Technology, Information and based on current standards for machine-readable descriptions of communications, Gwangju, Republic of Korea biological objects in SBML (MIRIAM-compliant RDF annotations with bioqualifiers) and uses a a large collection of synonyms and Objective: Organ Centric Protein Interaction Database (OCPID) database identifiers to match and compare model elements. The is a bioinformatics web tool, developed to provide integrated tool detects various kinds of syntactic and semantic conflicts analysis of human protein-protein interaction data and organ/ automatically and guides the user in resolving them. The tissue specific gene expression datasets. Although several underlying methods can also be integrated into other programs protein interaction databases collecting small-scale and large for model analysis. We have used them, for instance, to cluster scale experiments are currently available, they do not provide all models from the biomodels database based on common tissue-specific protein interactions which are valuable information substances appearing in the different models. for biological research such as finding tissue-specific interacting Conclusion: SemanticSBML is the first software tool for merging proteins of transcription factors with high reliability. of SBML models. It supports all relevant standards and provides Results: OCPID collects human protein interaction data sets a simple and intuitive way of recognising and resolving conflicts from DIP, HPRD, IntAct, and MINT. The application re-structured during model merging. protein interactions to visualize protein interaction network as function of organ/tissue. OCPID integrates organ/tissue specific gene expression data sets from several experiments, resulting in over 20,000 genes for 250 samples from more than 35 different organs. Organ/tissue specific genes are identified using Significant Analysis of Microarrays (SAM). Conclusions: Web interface includes search tools to allow users to query protein interactions with inputs of organs and a protein

184 ICSB 2008 in several different protein identifier schemes. It is also provided provided the best results for all the problems, both in terms of to search organ/tissue specific gene expression levels of genes final cost function value and least computational effort. in more than 35 different organs. The query result is enriched by Conclusions: A new global optimization method for parameter including protein domain and protein function information such estimation of nonlinear dynamic biological systems is presented as Pfam and Gene Ontology. In addition, network visualization in this study. The algorithm was interfaced with the Systems of protein interactions facilitates the global view of protein Biology Toolbox 2 for Matlab and proved to be highly effective interactions in an organ/tissue, and makes a novel discovery for solving a set of challenging parameter estimation problems, possible through visual inspection of network. outperforming all the other solvers available in this toolbox. Remark: The benchmark problems tested in this work are DS3-3-27 available in the toolbox, allowing interested users to try their own favorite optimization methods. JWS online web services: A tool for the programmatic References: query and simulation of biochemical pathway models Egea, J. A., Rodríguez-Fernández, M., Banga, J. R., Martí, R. van Gend, Carel; Snoep, Jacky Journal of Global Optimization. 2007, 37(3): 481-503. University of Stellenbosch, Biochemistry, Stellenbosch, South Rodríguez-Fernández, M., Egea, J. A., Banga, J. R. BMC Africa Bioinformatics. 2006, 7: 483. Schmidt, H., M. Jirstrand . Bioinformatics. 2006, 22(4): 514-515 Objective: JWS Online has long provided a means for performing (www.sbtoolbox2.org). Web-based simulations of biochemical pathway models. Using a web browser, users may search for and query models in the DS3-3-29 JWS Online database, and alter model parameters and launch simulations from a Java applet in the browser. Although this is COPASIWS - simulation web services with COPASI convenient for many users, there are many situations in which the Dada, Joseph1; Mendes, Pedro2 results of JWS Online simulations are required as input for other 1University of Manchester, Manchester Interdisciplinary Biocentre, analysis tools. To enable this, we have set out to create a suite of Manchester, United Kingdom; 2University of Manchester, School web services which allow programmatic access to JWS Online of Computer Science, Manchester, United Kingdom and the embedding of its functionality in workflows. Results: We have implemented the web services as Java code Objective: COPASI is a software tool that provides easy access using the Apache Axis framework. This makes development to powerful numerical methods for simulation and analysis of and extension of functionality straightforward. Available web biochemical reaction networks. CopasiSE, the command line services are documented in the Web Service Description version does not support model editing but can easily be used to Posters Language (WSDL) file which is automatically generated by the run simulations in batch mode. In this paper, we present work that Dedicated Axis framework. Service requests are transmitted as Simple exposes the powerful functionalities of COPASI as Web Services Object Access Protocol (SOAP) encoded XML messages, with using the command line version. Our main objective is to enable the arguments required for a particular request detailed by the flexible integration of COPASI functionalities with other local and WSDL file. Web service responses are likewise returned as SOAP remote services. By using Web Services, COPASI capabilities encoded XML. can easily be assessed in a language and platform independent Conclusion: Users request a web service and receive the results manner. Above all this will provide COPASI functionality for by means of XML-encoded messages, transmitted over HTTP. distributed computing. Following this protocol, users can query the database, obtain Results: We have implemented a prototype of the CopasiWS specific model information, and perform the various kinds of that runs inside an Apache Axis2 engine in a Tomcat Servlet simulations offered by JWS Online, and embed the results in a container. Possible clients are workflow engines, other Web workflow involving repeated calls to JWS Online or other web Services, Internet Portal or even CopasiUI running on user’s services. The simplicity and public availability of the SOAP and desktop computer. One of the benefits of this implementation is WSDL specifications means that it is easy to develop applications that it provides a flexible way to integrate different SBML models which utilize web services. A number of such applications exist, and simulation methods through a workflow engine. CopasiWS and we illustrate our application using Mathematica and Taverna produces output of a simulation in either COPASI native to create workflows consuming JWS Online and other, third-party output format or in XML format (SBRML, that we are presently web services. developing). Conclusions: CopasiWS is composed of several synchronous DS3-3-28 or asynchronous Web Services. The synchronous services are used for the running simulations that take short time to A new tool for parameter estimation in nonlinear dynamic run while asynchronous services are used for those that take biological systems using global optimization longer run times. In both cases, the user can supply the model Egea, Jose A.1; Schmidt, Henning2; Banga, Julio R.1 in either SBML or CopasiML formats together with simulation 1I.I.M. (C.S.I.C.), Process Engineering Group, Vigo, Spain; parameters. Each COPASI task, e.g. Time Course, is exposed 2University of Rostock, Systems Biology and Bioinformatics as a Web Service, and each of these has a specific set of control Group, Rostock, Germany parameters. Each Web Service interface is described by a specific WSDL. The system is currently available only as a prototype. Objective: Estimating parameters in nonlinear dynamic biological Public access is expected soon. models is a difficult task that often requires the use of global optimization techniques to avoid convergence to local solutions DS3-3-30 or to surmount some difficulties such as over-determined or badly scaled models (which results in very flat objective function areas). Optimization method of simultaneous equations in The objective of this work has been to develop an efficient and biological models using graph theory robust global optimization tool for reliably solving this class of Shimayoshi, Takao1; Amano, Akira2; Matsuda, Tetsuya2 problems, and to implement it in combination with a user friendly 1ASTEM Research Institute of Kyoto, Kyoto, Japan; 2Kyoto software toolbox. University, Graduate School of Informatics, Kyoto, Japan Results: We have extended and improved our global optimization algorithm SSm, based on scatter search (Rodríguez-Fernández Objective: Mathematical models on biological systems such et al., 2006; Egea et al., 2007) and have interfaced it with the as signal transduction can be formulated as ordinary differential Systems Biology Toolbox 2 for Matlab (Schmidt and Jirstrand, equations in conjunction with simultaneous equations, which 2006). A set of benchmark problems was solved with our new appear under instant equilibrium assumption for reactions. By algorithm, fSSm, and other 13 local and global optimization straightforward translation of equilibrium reactions to equations, algorithms available in the toolbox. Our method consistently each system of simultaneous equations has high degree. In

ICSB 2008 185 this study, an optimization method is introduced to reduce the DS3-3-32 degrees of simultaneous equation systems. In the method, the degree of a simultaneous equation system can be reduced A link between multiple high throughput technology by algebraic transformations of constitutive equations. The datasets transformations are found using graph algorithms on a graph Bucher, Elmar1; Kohonen, Pekka1; Mpindi, John-Patrick2; Sara, representation of the equation system. Henri2; Pisto, Tommi1; Fey, Vidal1; Edgren, Henrik3; Kilpinen, Results: The proposed method was applied to a model on Sami3; Ojala, Kalle3; Vuoriluoto, Karoliina2; Vainio, Paula1; Iljin, β-adrenergic control [Saucerman et al., J Bio Chem, 2003], which Kristiina1; Sahlberg, Niko1; Rantala, Juha1; Makela, Rami1; contains five systems of simultaneous equations. As a result, Saviranta, Petri1; Perala, Merja1; Nees, Matthias1; Haapa- the square sum of all degrees, which is an index of numerical Paananen, Saija1; Kallioniemi, Olli3 calculation time to solve the systems, became 11% of the original 1VTT Technical Research Centre of Finland, Medical value. Time to simulate the model for ten seconds in 0.01 ms Biotechnology, Turku, Finland; 2University of Turku, Turku, Finland; steps using the Euler and Newton method is reduced to 64.9% 3University of Helsinki, FIMM, Institute of Molecular Medicine for on average. In addition, average CPU time for the proposed Finland, Helsinki, Finland optimization method was 2.0 ms. Conclusion: The proposed method can considerably reduce Objective: Our laboratory performs high-throughput screening the degree of a simultaneous equation system, in addition to (HTS) experiments using siRNA, miRNA and compound libraries small processing time for the proposed algorithm. The efficiency on cancer-cell lines. We utilize three technologies: plate-based of simulations using models with simultaneous equations can be screens with different assays as endpoints, transfected cell arrays certainly improved. The proposed method will facilitate simulation with up to 5 antibody endpoints, and protein lysate arrays with study on biological systems. many immunostaining endpoints. In addition, we analyze samples by three commercially available microarray technologies: gene DS3-3-31 expression by Affymetrix arrays, miRNA expression by Agilent arrays, and gene copy numbers by Agilent arrays. This enables us COTRASIF: Genomics tool for systems biology to study the same biological sample from multiple angles, which Tokovenko, Bogdan1; Golda, Rostyslav2; Protas, Oleksiy2; provides a unique integrated opportunity to understand cancer. Dedicated Obolenskaya, Maria1 In order to accomplish such integrated systems biology -style Posters 1IMBG, Genetic Information Translation Mechanisms Dept, Kyiv, analyses on cancer, a database concept is needed. Ukraine; 2NaUKMA, Kyiv, Ukraine Results: Our database-design is structured in 3 parts: 1) HTS screen design, microarray and biological annotations, 2) reagent Objective: Promoter analysis and TFBS (transcription factor library annotations, 3) raw and normalized results from HTS. binding site) identification are essential for the understanding Annotations were available in Excel files. Raw-data were available of gene regulatory networks. Increasing specificity of the TFBS as csv-text-files from the plate reader or in the common file- prediction in eukaryotic gene promoters is a challenging task for formats from the microarray manufactures. modern bioinformatics. Database implementation was done using MySQL code. Upload, Based on our previous research, we observed better specificity normalization, and linking of the reagent-identifiers to Ensembl of the TFBS search when comparing promoters of orthologous genes was done using R scripts. genes of the evolutionary close species (e.g. rat and mouse) for An R library was written to access the database content, utilizing the presence of the target TFBS. the expression set class concept, a Bioconductor standard. This Our aim was to develop an easy-to-use web-tool for genome- gave us access to a repository of free open source software, wide identification of putative TFBS with enhanced results quality. developed for the analysis and comprehension of genomic data, Results: We developed a new tool (COTRASIF, conservation- which can be run, studied, modified, and redistributed as needed. aided transcription factor binding site finder) for the genome-wide Conclusions: Our integrated high-throughput database has now identification with increased specificity of the putative TFBS in made it possible to study our data in completely new ways. To eukaryotic gene promoters. show the utility of the integrated database concept, we developed COTRASIF is built upon the semi-automatic importer of ways to produce e.g. different plots for individual genes across all promoters from the Ensembl genome annotation database. the different experiments carried out with different technologies. Currently COTRASIF has 11 genomes available. Promoters are defined as 800bp upstream from transcription DS3-3-33 start site, plus the 5’ UTR coding sequence. For TFBS search, both position-weight matrix (PWM) approach and the recently Novel network analysis tools to characterize Bacillus developed HMM-based (hidden Markov models) approach can subtilis mutants with alternative catabolic pathways be used. Search results can be further analyzed using the built- Florez, Lope A; Gunka, Katrin; Commichau, Fabian M; Stuelke, in gene orthology filter. Orthology information is automatically Joerg obtained from the Ensembl Compara genome alignments Institute of Microbiology and Genetics, University of Gottingen, database. Department of General Microbiology, Gottingen, Germany Further development includes: addition of new genomes; integration of the Gene Ontology category enrichment functional Objective: In the soil bacterium Bacillus subtilis the catabolism analysis (hypergeometric and Bayesian); more results output of amino acids occurs primarily by the glutamate dehydrogenase formats; specialized web-API (application programming interface) RocG. Alternatively, the cryptic glutamate dehydrogenase for enabling easy use of COTRASIF by other tools. GudB is activated in rocG mutants. We knocked out the genes Conclusions: We identified 323 genes in rat genome which encoding RocG and GudB. The double mutants are unable to contain ISRE (interferon-stimulated response element) in their grow in a medium that contains glutamate as the only carbon promoters, and are the potential targets of transcriptional source. Interestingly, suppressor mutants were isolated from this regulation by type I interferons. Functional analysis of these strain that can utilize glutamate. Implementing a combination of genes, conducted using Gene Ontology Tree Machine, had network analysis algorithms we wanted to characterize the newly shown the relative enrichment of 24 GO categories, of which 12 acquired suppressor mutation. represent known IFN effects. Results: We developed two algorithms that use the current Availability: COTRASIF is freely available at http://biomed.org. genome scale metabolic model of B. subtilis [1] as input. The ua/COTRASIF/ algorithms were designed to find potential degradation pathways and rank them based on these criteria: they should start with glutamate and lead to a substrate of the central metabolism; the required substrates for each reaction should be available; and the number of required mutations should be minimized (Occam’s razor). The first algorithm assumes that no other starting

186 ICSB 2008 metabolites are present for catalysis apart from glutamate. The Results: MatCont is an interactive Matlab software package second algorithm assumes a starting pool of other metabolites. for the nummerical study of continuous (ODE-based) dynamical Each result of the latter algorithm is a combination of pathways systems. It is basically a continuation package (like AUTO that yields a net gain of metabolites after glutamate degradation. and CONTENT), i.e. its kernel is a continuation routine for Conclusions: The first algorithm shows that — if we exclude the computation of solutions to a system of equations under RocG and GudB — it is not possible to catabolise glutamate parameter variation. The system of equations can describe unless we assume that additional metabolites are present in the any type of objects, for example equilibria, limit cycles, Hopf cell as well. These additional metabolites are required in the first bifurcation points, bifurcations of limit cycles or homoclinic orbits. steps of catalysis and are replenished later. The second algorithm The number of free parameters must be one higher than the was able to find potential catabolic pathways under this condition. codimension of the bifurcation, since we need systems where A ranking of the results indicates that the genes involved in the number of variables is the number of equations plus one. aspartate degradation are likely involved in the process. This is During continuation of codimension 0 or 1 objects, all generically confirmed by wet lab experiments. possible bifurcations of higher codimension are detected, located [1] - Oh et al. (2007), J. Biol. Chem., 282:28791-28799. and their normal forms are computed. Furthermore, from a codimension 1 or codimension 2 bifurcation point it is possible to DS3-3-35 start the curves of lower codimension that are generically rooted in such points. SymCA: A generic implementation of the metabolic control Conclusions: The bifurcation behaviour of cell cycle control can analysis control-matrix equation be studied in an interactive and user-friendly way by using the Akhurst, Tim; Hofmeyr, Jan-Hendrik; Rohwer, Johann Matlab package MatCont. It has functionalities that cannot be Stellenbosch University, Biochemistry, Stellenbosch, South Africa found in any other software.

Objective: Metabolic Control Analysis (MCA) provides a powerful DS3-3-37 quantitative framework for understanding and explaining the relationships between the steady-state properties of a cellular Protein and domain interaction inference in the switching system and the individual components comprising the system. motility system of Myxococcus xanthus using Cytoprophet Traditionally, both MCA and kinetic modelling have focussed Morcos, Faruck1; Sikora, Marcin2; Lamanna, Charles1; Alber, on numerical solutions, with algebraic solutions being tackled Mark3; Kaiser, Dale4; Izaguirre, Jesús1 less frequently. Our objective is thus to develop a generic 1University of Notre Dame, Computer Science and Engineering, implementation that will algebraically express the control Notre Dame, United States; 2University of Notre Dame, Electrical 3 coefficients of MCA in terms of elasticities; this approach is based Engineering, Notre Dame, United States; University of Notre Posters on the control-matrix equation illustrated in [1]. Dame, Mathematics, Notre Dame, United States; 4Stanford Dedicated Results: We here present SymCA, a Python add-on module for University, Stanford, United States the PySCeS simulation software, which can derive these algebraic relations for systems of any size and complexity. The Maxima Objective: We introduce a belief propagation method to infer software provides the symbolic maths environment, and to this protein and domain interactions in Myxococcus xanthus. This end a Python interface was developed using the subprocess approach creates a factor graph representation of the joint module. Initial structural analysis of the stoichiometric matrix is probability distribution (JPD) of protein and domain interaction performed using PySCeS, and all subsequent computations are partners with experimental interactions, domain architecture and handled via SymCA. We have successfully tested SymCA with a GO annotations. We use complexed structures coming from number of biological models from the JWS online database by experiments or computational docking as a priori evidence of computing the control coefficient expressions and substituting the domain interactions. steady-state flux, concentration and elasticity values, and finally Results: The JPD of protein and domain pairs is marginalized comparing the results to numerical simulations with PySCeS. efficiently with the Sum-Product Algorithm. This framework has Conclusions: The combined power of Python, PySCeS and been implemented in a software plug-in for Cytoscape that we Maxima has resulted in the successful implementation of the called Cytoprophet. It implements three algorithms that predict control-matrix equation in a new software tool, SymCA, which new potential physical interactions. The algorithms for protein can perform symbolic control analysis on biological systems of and domain interaction inference include Maximum Likelihood ranging size and complexity. We are now able to quantify the Estimation (MLE) using Expectation Maximization (EM); the set significance and influence of each control pattern [2] within a cover approach Maximum Specificity Set Cover (MSSC) and the control coefficient expression. Such control patterns can be Sum-Product Algorithm (SPA). Cytoprophet draws a network visualised as “routes of regulation”; their quantification may thus of potential protein and domain interactions with probability lead to identification of key regulatory pathways, which could be scores and GO distances as edge attributes for a set of input applied e.g. in drug discovery. proteins with domain annotation. We use SPA algorithm within References: 1.Hofmeyr, J-H. (2001). Proceedings of 2nd Cytoprophet to study Myxococcus xanthus reversal system. We International Conference on Systems Biology. 291 infer interactions in a signaling pathway that regulates motility. 2.Hofmeyr, J.-H. (1989). Eur. J. Biochem. 186, 343 Conclusions: Results show agreement with experimental results and support a series of potential interactions involving the mutual DS3-3-36 gliding protein MglA. This protein has been hypothesized to bridge a biochemical oscillator that synchronizes reversals and the Numerical bifurcation study of a cell cycle model by Tyson machinery that constructs motility engines. A potential interaction and Novak between the Ras domain in MglA and the response regulator Govaerts, Willy; Kheibarshekan, Leila in protein FrzE of the motility synchronization circuit provides Ghent University, Applied Mathematics and Computer Science, computational evidence that this is possible. Cytoprophet Ghent, Belgium software, documentation and sample sessions are freely available at http://cytoprophet.cse.nd.edu. Objective: Fundamental work in the mathematical bifurcation approach to the study of cell cycle control was done by J.J. Tyson and B. Novak, see e.g. Chapter 10 in “Computational Cell Biology”, Springer 2002. This involves saddle-node bifurcations, Hopf bifurcations, stable and unstable limit cycles, homoclinic-to- saddle node orbits and two-parameter bifurcation diagrams. Except in the simplest cases, such models can only be studied by specialized software such as AUTO, CONTENT or MatCont. We discuss the user-friendly package MatCont for this purpose.

ICSB 2008 187 DS3-3-38 Conclusions: We have developed KEGG2SBML to convert KEGG metabolic pathway database into SBML files. With this User extensible and customizable integrated management converter, we have succeeded in converting about 60,000 framework for biological data KEGG metabolic pathways. Existing SBML-aware applications Maeshiro, Tetsuya; Nakayama, Shin-ichi can directly use these converted pathways. KEGG2SBML is University of Tsukuba, School of Library and Information Science, available from http://sbml.org/kegg2sbml.html as an open-source Tsukuba, Japan package.

Objective: We are developing a web based system for uniform DS3-3-40 management of biological data, such as gene sequences and their locations in genome, protein sequences, structures and An integrative approach to modeling in systems biology fuctions, gene regulatory networks, several types of images, and Bergmann, Frank; Sauro, Herbert results of high throughput experiments. Conventional systems University of Washington, Bioengineering, Seattle, United States and web sites are designed by the creators of the corresponding systems, and these are user friendly for the creators. However, Computational approaches have a strong influence on biology: not everyone feels confortable with the usability and the types of Next to established fields like Bioinformatics, new fields like integrated data of available systems and web sites. Our system Computational and Systems Biology have become established. allows users to customize the data presentation that suit best The knowledge and experienced learned in these fields have for their purpose and research environment, by choosing the been taken up by the scientific community and find applications data type to be shown and modifying the layout. Internal data in industry. representation is based on an extended graph representation Both CellML and SBML have emerged as de facto standards model, which has more flexibility to associate data than and are broadly supported by software applications. Their current conventional frameworks such as relational data model. development focuses primarily on improving model composition Results: The prototype of the system, that manage data related and enriching the encoded computational model by means of to C.elegans, is operational, which incorporates gene sequences, meta-information in form of various ontologies. However so protein sequences, structures and functions. Experimentally far hardly any software application makes use of these new Dedicated elucidated gene regulatory networks and signal transduction language features and prototypes and use-cases are sorely Posters networks, besides the microarray expression data, are also missing. Current modeling packages focus on specific aspects included into the system. Customization of keyword search result of modeling and dedicated simulation packages exist to simulate was provided to users in our institution for evaluation, and positive computational models. Even though modeling packages provide feedback was returned regarding system usability. rudimentary support for in-place simulation, as exemplified Conclusions: Customization of user interface and data by tools like JDesigner and CellDesigner, the feedback from management increases considerably the usability of data simulations is not immediate and rather disconnected from management. the overall modeling process. Researchers would benefit from a tighter coupling of the modeling process and simulation DS3-3-39 capabilities. This coupling would facilitate experimentation with the models in a way that was previously only possible using script KEGG2SBML: A tool for converting KEGG metabolic based modeling environments like Jarnac. A combination of pathway database to SBML modeling, simulation and visualization will make a new modeling Jouraku, Akiya1; Ohta, Nobuyuki1; Funahashi, Akira1; Kitano, tool a valuable asset in teaching and collaboration efforts. Hiroaki2 This poster presents a new integrative modeling environment, 1Keio University, Department of Biosciences and Informatics, which combines common modeling metaphors with continuous Yokohama, Japan; 2The Systems Biology Institute, Tokyo, Japan simulation. Simulation will be included in the modeling process with novel visual aids for specifying kinetics and discrete events to Objective: Identification of biochemical networks by using be included in computational models. A dedicated runtime view computational modeling is one of the major challenges of systems will provide an ideal experimentation test-bed for computational biology. For computational modeling, the Systems Biology models (i.e.: by providing facilities for knockout experiments). Markup Language (SBML) has been widely used to represent and The new modeling framework will provide several Extensibility exchange the biochemical reaction networks based on a common modes, for the application to be enhanced by dedicated plug-ins, XML format. Several dozen simulation, analysis, and modeling modules for the Systems Biology Workbench and custom scripts. applications have supported SBML and more in the process The project is still in progress. of being extended to support it. On the other hand, several efforts have been made to create large-scale, comprehensive DS3-3-42 databases of biochemical networks based on literatures. Kyoto Encyclopedia of Genes and Genomes (KEGG) is one of the Simulating the reaction-diffusion master equation on most widely used databases. More than 70,000 pathways for unstructured meshes more than 700 organisms (in Apr. 2008) are provided in KEGG Hellander, Andreas; Engblom, Stefan; Ferm, Lars; Lötstedt, Per PATHWAY database. Making these pathways available in SBML Uppsala University, Information Technology, Uppsala, Sweden format is quite useful by the following reasons: (1) it will enable researchers to apply many useful SBML-aware applications to Objective: Stochastic simulation of reaction-diffusion processes these pathways, and (2) this leads to the valuable feedbacks for inside living cells are computationally very expensive. We extend the continued evolution of SBML. an existing algorithm in two ways. First, by making connections Results: We have developed KEGG2SBML to convert to the finite element method (FEM) we are able to conduct KEGG metabolic pathways into SBML files. KEGG2SBML is simulations on unstructured meshes. The jump propensities can implemented as a Perl script and thus can be used in various be conveniently obtained by using existing FEM software, and platforms. KEGG2SBML can convert the latest release (Release thus the treatment of complicated geometries is made possible. 46.0 in Apr. 2008) of KEGG metabolic pathways into all Secondly, in the spatially homogeneous case, hybrid methods specifications of SBML up to the latest version of Level 2 Version have been introduced to facilitate the study of stiff models. 3. Layout information in KEGG pathway diagram can also be We propose a hybrid method for the reaction-diffusion master added in the converted SBML (up to Level 2 Version 1). The equation where the diffusion is treated deterministically, when layout information is stored in SBML annotations and can be appropriate, for some or all of the species. The connection to used in CellDesigner, a process diagram editor for biochemical FEM has the potential of making this approach highly efficient by networks developed by us. We have succeeded in converting utilizing the capabilities of state of the art solvers. about 60,000 pathways from the latest release of KEGG Results: We show that the method qualitatively reproduces metabolic pathway database. results obtained with an existing software (mesoRD) that uses

188 ICSB 2008 Cartesian meshes. In a first Matlab implementation, the hybrid directly using IDA. KINSOL uses Krylov methods to solve approach is three orders of magnitude faster for a model problem. nonlinear systems, allowing for the solution of larger systems Conclusions: The results suggest that the method will be an than the direct methods used by HYBRD and NLEQ2. CVODES efficient alternative to pure stochastic simulation. Ongoing work provides forward and adjoint sensitivity analysis capabilities. includes the integration of the method with finite element software Conclusions: Using PySUNDIALS, PySCeS, in addition to in order to obtain a generic, easy-to-use framework for both providing a greater number of alternative solvers, now implements stochastic and hybrid simulation of reaction-diffusion models. additional SMBL features, including events, a differential- algebraic equation framework, the application of algebraic and DS3-3-43 general constraints to variable compartment size modeling, and constraints. The standalone nature of the PySUNDIALS A Matlab toolbox for nonlinear mixed effects modeling - module makes it available for use to a wider scientific audience. illustrated with a model for lipoprotein kinetics PySUNDIALS is open-source software available under an MIT Sunnåker, Mikael1; Berglund, Martin2; Wennberg, Bernt2; Adiels, license from http://pysundials.sourceforge.net, and PySCeS is Martin3; Jirstrand, Mats1 available under the GPL from http://pysces.sourceforge.net. 1Fraunhofer-Chalmers Centre, Gothenburg, Sweden; 2Chalmers University of Technology, Mathematical Sciences, Gothenburg, DS3-3-45 Sweden; 3Gothenburg University, Wallenberg Laboratory for Cardiovascular Research, Gothenburg, Sweden “Payao”: Web community tagging system to SBML models Matsuoka, Yukiko1; Kikuchi, Norihiro2; Sugimura, Haruka1; Objective: Mathematical models constitute an important step Hayama, Akemi1; Kitano, Hiroaki3 towards an increased understanding of the processes underlying 1The Systems Biology Institute, Tokyo, Japan; 2Mitsui Knowledge biological systems. Such systems are often complex, which puts Industry Co., Ltd., Tokyo, Japan; 3Okinawa Institute of Science high demands on the modeling techniques used. A crucial part of and Technology, Okinawa, Japan any modeling effort is to estimate the model parameters, which are often uncertain and may even vary between subjects. These Objective: Modelling a biological pathway and network requires issues can be handled by applying a statistical framework known accurate curation and constant updates with the latest data. as nonlinear mixed effects (NLME) modeling, in combination with What is needed is a framework to facilitate tracking the latest stochastic differential equations (SDEs). information and update mechanism for the collaborated model Results: We present a Matlab toolbox that can be used for building. To provide such a framework is essential for systems model specification, simulation, and parameter estimationof biology research.

NLME models with SDEs. The toolbox, which is called NLMEtools, Results: We have developed the system called “Payao”. “Payao” Posters is based on the Systems Biology Toolbox for Matlab which has aims to enable a community to work on the same models Dedicated previously been developed at the Fraunhofer-Chalmers Centre. The concurrently, and share the up-to-date knowledge among them. toolbox constitutes a natural platform for the analysis of biological The uesrs can insert tags to the specific parts of the model, data from several individuals. A user-friendly graphical user exchange comments, search for the relevant publication, archive interface is provided to support the system identification workflow the discussions so that they can update the models accurately. from specification of the model structure to estimating parameters, “Payao” reads the models in Systems Biology Markup Language as well as generation of synthetic data and comparison with real (http://sbml.org) format, displays them with CellDesigner, a estimation data. process diagram editor (http://celldesigner.org), adopting the Conclusions: The parameters of a previously published Systems Biology Graphical Notation (http://sbgn.org), and compartmental model for lipoprotein kinetics in the blood plasma provides an interface for adding tags and comments to the are analyzed, as an illustrative example of the usage of the toolbox. models for the community members. To monitor the kinetics of the lipoproteins an isotope-labeled tracer, The system consists of client / server utilities: Payao server uses leucine, is used in the experimental setup. The kinetic parameters CellDesigner 3.5.1 plugin API and obtains the model information of the model are estimated for artificially generated data and the to display. Users can add information such as keywords, links, statistical properties of the estimates are discussed. pubmedID and free text as tags to specified components in the model. Other users can contribute comments to the tags, It DS3-3-44 maintains the user list with access controls, so that the models can be properly managed. The information on the login users and PySUNDIALS: New capabilities and alternatives for Python tags/comments data are stored in RDBMS. The client tool is build based systems biology software by FLEX system. Dominy, James; Olivier, Brett G; Rohwer, Johann M; Hofmeyr, Conclusions: We have developed a web-community tagging Jan-Hendrik S system called “Payao”, which set SBML models and tag them on Stellenbosch University, Biochemistry, Stellenbosch, South Africa the web by the community members. The system aims to be a innovative model sharing framework, and serve as a knowledge Objective: The Python Simulator for Cellular Systems (PySCeS) base for systems biology research. The website for Payao is is a Python module which allows interactive, programmatic http://payaologue.org. modeling of cellular systems, and supports the Systems Biology Markup Language (SBML) for model interchange. Its functionality DS3-3-46 has been extended using its recently developed modular framework, by developing PySUNDIALS; a separate Python ModelMage: A tool for the automatic generation and module which is a high level wrapper for the SUite of Non-Linear discrimination of SBML-models DIfferential/ALgebraic Solvers (SUNDIALS). SUNDIALS is a Flöttmann, Max; Schaber, Jörg; Klipp, Edda flexible, feature rich, modern suite of solver routines capable of MPI for molecular genetics, Berlin, Germany running in both serial and parallel, which compares favourably to other software suites such as minpack, which has fewer Objective: Mathematical modeling of biological systems features, and lapack, which is focused explicitly on parallel involves implementation, testing and discrimination between implementations. model alternatives that differ in the number of components, Results: PySUNDIALS provides Python bindings to the complete reactions and/or kinetics. Generating and managing these model SUNDIALS suite, which is written in C. The Python built in module alternatives is a tedious and difficult task and can easily lead to ctypes is used for the actual foreign function interface. The errors. ModelMage is a management tool that facilitates handling individual solvers of SUNDIALS facilitate the provision of various of candidate models. It is designed for the easy and rapid new features within PySCeS. CVODE is used to support SBML development, generation, simulation and discrimination of model events and constraints using root finding. SBML algebraic rules, alternatives. such as those specifying moiety conservations, are supported Results: The main idea of the program is to create a defined set

ICSB 2008 189 of model alternatives in an automatic way. The user provides only wide variety of problems. one SBML-model and a set of directives from which alternatives Here we present a software toolbox, DOTcvp, which uses the are created by leaving out components and/or reaction. After CVP approach for handling continuous and mixed integer DO generating the models, the software can automatically fit all these problems. DOTcvp has been successfully applied to several models to data and provide different statistical measures for problems in systems biology and bioprocess engineering. The goodness of fit to make discrimination between the models easier. toolbox is written in MATLAB and provides an easy to use In contrast to other model generation programs, ModelMage aims environment while maintaining a quite good performance. DOTcvp at generating only a limited set of models instead of all possible is designed for the Windows and Linux operating systems. The ones. Moreover, it uses COPASI as a simulation engine. Thus, toolbox also contains a function for importing SBML models. The all simulation and optimization features of COPASI are readily code and the documentation, with many example problems, is incorporated. available at: Conclusions: With ModelMage we found a way to quickly http://www.iim.csic.es/~dotcvp/ generate families of models and rank them by goodness of fit to Results and conclusions: In this contribution, we present given data. During tests with artificial as well as real world models the most relevant aspects of the DOTcvp software toolbox and from the literature, ModelMage proved as a useful tool that could its application to several challenging problems, like the phase speed up modelling work substantially in some cases. resetting of a calcium oscillator, which is characterized by instabilities of the steady states. The solution of this and other DS3-3-47 DO problems reveal DOTcvp as a very promising tool in systems biology. BlastXtract2 - Improving the early exploration of (meta) genomic sequences by intuitive visualization and DS3-3-49 management Claesson, Marcus; O’Toole, Paul Arcadia: a visualisation tool for metabolic pathways University College Cork, Microbiology, Cork, Ireland Villeger, Alice1; Pettifer, Steve2; Kell, Douglas1 1The University of Manchester, School of Chemistry, Manchester, Objective: The ever-increasing rate with which genomic data is United Kingdom; 2The University of Manchester, School of Dedicated generated is unfortunately not matched by sufficient progress in Computer Science, Manchester, United Kingdom Posters bioinformatics applications capable of handling these enormous amounts of sequence data. In particular, the recent introduction Objective: As the amount of data available on biological of next-generation sequencing technologies has accelerated systems increases, so does the need for computing tools this development, at the same time as it provides us with supporting their analysis. Notably, visualisation software letting opportunities to sequence more genomes faster, and to a lower scientists interactively explore various kind of pathways provide cost. This also applies to metagenomes, where partial genomes considerable assistance in making sense of complex networks. of a multitude of microbial organisms from environmental samples Unfortunately, popular existing applications (such as Cytoscape) are assembled and analysed. struggle to address adequately the specific case of metabolic Results: The first version of BlastXtract was used to visualize and pathways: automatic layouts computed by generic graph-drawing manage all-ready performed translated BLAST or FastA searches algorithms usually require time-consuming manual adjustments of single sequences against a single protein database at a time. from the user. Consequently, we designed a piece of software In this second generation of BlastXtract, BLAST searches can be specialising in the representation of metabolic pathway. launched from within the program, against an arbitrary number of Results: Written in C++, Arcadia makes use of powerful sequence databases. Moreover, translated searches of multiple existing open-source libraries: LibSBML provides an interface to sequences can now be simultaneously displayed, one query SBML pathway files, the Boost Graph Library an efficient graph sequence after another. Some additional enhancements include structure, GraphViz classic layout algorithms, libavoid dynamic easier display of best-hits in every non-overlapping position, edge routing, and Qt the Graphical User Interface. Cross- integration of external gene-predictions, and the option to extract platform, the code can be compiled to run on Mac, UNIX or and save amino acid sequences of any relevant hits. Windows systems. Our tool enables navigation between multiple Conclusions: Collectively, these improvements make interconnected views of the same model: sorted lists of reactions BlastXtract2 the ideal tool for early sequence- similarity-based and biochemical species, detailed properties, global map, groups exploration of large nucleotide sequence batches, either in draft of neighbouring molecules. Intuitive, context-sensitive mouse- or final versions, generated by single- or meta-genomic projects. controls let the user apply local layout strategies to particular subsets of the network. Default behaviours have been defined for DS3-3-48 typical domain-specific concepts such as “backbone”, “modifier” or “cofactor”, resulting in a semi-automated layout. Additionally, DOTcvp - a software toolbox for dynamic optimization in metadata annotations identifying different types of reactions and systems biology species are used to define their visual appearance, in accordance Hirmajer, Tomas; Balsa-Canto, Eva; R. Banga, Julio with the SBGN recommendations. IIM-CSIC, Vigo, Spain Conclusions: Our prototype opens the way to further developments: namely, fully automated layout driven by semantic Objective: Optimization aims to make a system or design information extracted from meaningful patterns (e.g. cycles) as effective or functional as possible. In systems biology, detected through graph analysis. Functional extensions are most models have a dynamic nature, usually consisting of also planned through the Utopia and Taverna frameworks, for sets of differential equations. Dynamic optimization seeks the a seamless integration of online tools (e.g. text mining) and computation of optimal time-varying conditions (called control resources. Finally, support of the SBRML format would enable the variables) for this type of systems. These control variables allow graphical representation of simulation results onto the pathway. us to stimulate a given biosystem so as to achieve a desired dynamic behavior, e.g. stimulate a cell signaling cascade to maximize amplitude strength and/or duration, stimulate an oscillatory system to synchronize oscillations, remove oscillations, etc. In other words, DO allows the computation of optimal operating policies for processes or systems which maximize (or minimize) a pre-defined performance index subject to the system dynamics and other possible constraints. DO may be solved by several methods, including the control vector parameterization (CVP) method, which has demonstrated excellent performance for a

190 ICSB 2008 DS3-3-50 5000 confocal images of segmentation gene expression patterns in Drosophila. The constructed dataset has cellular resolution The SBML test suite in space, 6.5 minutes resolution in time and spans about 1.5 Keating, Sarah1; Begley, Kimberly2; Hucka, Michael3 hours of development (Surkova et al., 2007). All the images and 1California Pasadena Institute of Technology, Engineering and quantitative data are stored in the FlyEx database (http://urchin. Applied Science, Pasadena, United States; 2Finance Science spbcas.ru, http://flyex.ams.sunysb.edu/flyex) and are widely used Consulting Pty Ltd, Eagle Heights Queensland, Australia; by biologists to study the mechanism of pattern formation, infer 3California Institute of Technology, Engineering and Applied regulatory interactions in the segmentation genetic network and Science, Pasadena, California, United States develop new mathematical models (for references see http:// urchin.spbcas.ru/flyex/refs.jsp). The ProStack package was also Objective: A missing ingredient in the Systems Biology Markup successfully used to acquire quantitative data on the expression Language (SBML; http://sbml.org) community has been an official of other Drosophila genes, as well as data on gene expression in set of objective criteria for assessing a software tool’s degree of early development of the coral Acropora millepora and the sea coverage and conformance to SBML. Although the language anemone Nematostella vectensis. specifications define the syntax and semantics of SBML, it This work is supported by NIH grant RR07801, GAP award has remained a challenge for software developers to test for RUB1-1578, NWO-RFBR project 047.011.2004.013, and RFBR correct implementations. Further, software users have had no grants 08-01-00315a, 08-04-00712a. straightforward, independent means of checking for themselves how fully a given package conforms to the SBML standard. DS3-3-52 Results: We have now developed a full conformance testing suite for SBML. Its core is a corpus of several thousand hand- PathText/payao development: Text/pathway annotations written models, each of which tests different aspects of SBML Sætre, Rune1; Kemper, Brian1; Oda, Kanae1; Matsuoka, Yukiko2; syntax or semantics. Test cases range in complexity from trivial Kikuchi, Norihiro3; Kitano, Hiroaki4; Ananiadou, Sophia5; Tsujii, to advanced, including combinatorial tests for verifying multiple Junichi1 features working together. Every test model is accompanied by 1University of Tokyo, Computer Science, Tokyo, Japan; 2The human-readable descriptions, and (where appropriate) expected Systems Biology Institute, Tokyo, Japan; 3Mitsui Knowledge simulation results for specific simulation conditions. Every case Industry Co., Ltd., Tokyo, Japan; 4Okinawa Institute of Science enumerates machine-readable tags categorizing it along different and Technology, Okinawa, Japan; 5National Centre for Text dimensions; these tags are used by the test automation facilities. Mining (NaCTeM), University of Manchester, Manchester, United The SBML Test Suite is available as both a web-based facility Kingdom hosted at http://sbml.org and a standalone application for running Posters on a user’s personal computer. The software is written in Java, Motivation: The PathText project goal is to discover publications Dedicated portable to Linux, MacOS and Windows, and distributed for free relevant to entities in Pathway models, by using pathway under LGPL terms. Both the online and standalone systems allow visualization tools as a user interface, and various text mining the user to select which tests are performed based on the SBML tools to do “context sensitive” knowledge discovery. levels/versions, SBML components, and other attributes and Background: Most biomolecular knowledge is published as constructs from the SBML specification. As each test case runs, English text, for example in the PubMed database, but at the the system reports the pass/fail status; details are available at the same time higher level collective knowledge is often published click of a button. using a graphical notation (like SBGN) to represent all entities Conclusions: Software developers will greatly benefit from and their interactions in a pathway. We believe that such having a conformance testing suite to guide their SBML software pathway visualizations can serve as an effective user interface for development; likewise, users should find it easier to determine knowledge discovery, if they are linked to corresponding text in exactly which SBML features are supported by a given software journal publications. Since the graphical elements in a Pathway system. are of a very different nature than their corresponding descriptions in English text, we developed a prototype system called PathText, DS3-3-51 to serve as a bridge between these two different representations. Methods: We are now entering phase two of the PathText ProStack, an image analysis software for identification and project, and the next step is to create a bigger corpus of quantification of objects visualized with microscope annotated full text research papers, where each annotation in Kozlov, Konstantin1; Pisarev, Andrei1; Kaandorp, Jaap2; Gursky, a paper corresponds to a specific entity or reaction in a newly Vitaly3; Samsonova, Maria1; Reinitz, John4 created Pathway. This corpus can be used to train a computer 1St. Petersburg State Polytechnical University, St. Petersburg, system to add more similar annotations to the Pathway later, Russian Federation; 2University of Amsterdam, Section since new related papers are published every day. Computational Science, Faculty of Science, Amsterdam, The new corpus will also be an important resource for passive Netherlands; 3Ioffe Physico-Technical Institute, St. Petersburg, users, like biologists and others examining the Pathway, since Russian Federation; 4Stony Brook University, Stony Brook, NY, it provides textual evidence, establishing the correctness, and Russian Federation giving explanations for why specific Species and Relations in the Pathway were added by the original Pathway-creator. Objective: Biology is increasingly asking quantitative questions. Results: “PathText enabled” pathway models have been Quantification is essential to understand the principles of preprocessed to extract all information about the Species organism functioning. Modern physics and engineering provide (Entities), their Species-Aliases (Identifiers) and the relations tools and strategies for accurate measurements and the between them in the pathway. This information is now connected acquisition of comprehensive and consistent datasets. to existing Text Mining tools like MEDIE, Info-PubMed, FACTA and Results: We have developed a software package, known as KLEIO. ProStack, which implements many image processing operations, References: as well as all the methods required to quantitatively characterize http://www.nactem.ac.uk/pathtext/ objects visualized with microscope. ProStack was successfully http://www-tsujii.is.s.u-tokyo.ac.jp/medie/ applied to extract quantitative data on gene expression from https://www-tsujii.is.s.u-tokyo.ac.jp/info-pubmed/ images obtained with confocal microscope. http://text0.mib.man.ac.uk/software/facta/ Conclusions: ProStack allows the user to enhance image quality, outline regions occupied by cells and nuclei, measure the shape and dimensions of cells, calculate the statistical estimators of intensity inside cells or other compartments, visualize this information and store it as a text file or as a VRML model. The package was applied to extract quantitative data from about

ICSB 2008 191 DS3-3-53 -flux balance analysis -(elementary) conservation relations Metabolica - a tool for Bayesian analysis of metabolic -elementary modes and minimal cut sets networks The main features for signaling networks are: Heino, Jenni1; Calvetti, Daniela2; Somersalo, Erkki2 -computation of pos./neg. paths/cycles and minimal cut sets 1Helsinki University of Technology, Department of Mathematics -species interdependence and Systems Analysis, Espoo, Finland; 2Case Western Reserve -logical steady states and stimulus response analysis University, Department of Mathematics, Cleveland, United States -structural couplings of signal flows -strongly connected components Objective: Models for cellular level metabolism are typically -minimal intervention sets complex and lead to systems of high dimensionality. The CNA has a GUI which allows editing of the network structure and Bayesian framework is naturally suited to analyze complex and visualization of computation results. In its latest version (9.0), the potentially unstable systems and to infer on the state of the CNA follows a more object-oriented approach which makes it system using the scarce data and available prior information. possible to support API functionality that can be used to directly We have used the Bayesian methodology to analyze metabolic access several of the main calculation procedures. Metabolic networks and have set forth to publish an open source software networks can be imported and exported in SBML format. package called Metabolica that enables scientists without deep Conclusions: CellNetAnalyzer is a powerful analysis tool that has background in Bayesian analysis to utilize the methodology. been successfully applied to many networks so far. Results: The Metabolica package integrates tools for model construction, defining the Bayesian model, Markov Chain Monte DS3-3-55 Carlo (MCMC) sampling and visualization of the results. The model constructed is a spatially lumped multicompartment model StropE, the streptomyces operon prediction database for cellular level metabolism, where the compartmentalization server is done according to the physiology considering, e.g., different Laing, Emma1; Velarde, Giles2; Kell, Douglas2; Smith, Colin P1; intracellular organs or different cell types and extracellular space. Hubbard, Simon J3 The current version of Metabolica concentrates on the statistical 1University of Surrey, Faculty of Health and Medical Sciences, Dedicated approach to steady or stationary state inverse problem, where Guildford, United Kingdom; 2University of Manchester, School Posters the problem is to find the reaction and transport fluxes and blood of Chemistry, Manchester, United Kingdom; 3University of concentrations defining the state of the system. The Bayesian Manchester, Faculty of Life Sciences, Manchester, United approach implemented includes forming the posterior probability Kingdom density of the flux distribution given all available data and prior information, including bounds, Gaussian priors and beliefs on the Objective: Transcriptional regulation networks as a component state of the system and state uncertainties. The posterior density of an entire biological system are comprised of hierarchies of is explored using an efficient hybrid Gibbs sampler and Hit-and- regulation complexity that need to be considered individually; Run algorithm. Metabolica is implemented using MATLAB and is here it is the operon level being investigated. We have previously intended to be self-explanatory and easy-to-use. described how an operon can be predicted and what information Conclusion: The Bayesian framework and the MCMC (function/regulation) that the structure of the unit carries (Laing et sampling methods incorporated in Metabolica present a distinct al, 2006, 2007, 2008 and manuscript in preparation). However, approach to analyze the metabolic networks, and Metabolica will the increasingly common observation of operon expression complement the existing software tools in Systems Analysis. The dynamicity, the ability to change the regulation patterns of current release concentrates on the steady/ stationary state, and member genes under certain conditions, is a parameter that as the dynamic methodology is being developed for future releases. yet has been difficult to explore. Results: We have developed a tool, StropE, which allows DS3-3-54 the analysis of operon expression in Streptomyces coelicolor, an organism known to have highly complex transcriptional CellNetAnalyzer - structural and functional analysis of regulation. Taking a single or a combination of gene name(s) and biological networks user-selected microarray experiments (either publicly available Klamt, Steffen; von Kamp, Axel pre-loaded or loaded from http://strep-microarray.sbs.surrey. Max Planck Institute for Dynamics of Complex Technical Systems, ac.uk/maxdBrowse data or user’s own private data) as input the Systems Biology, Magdeburg, Germany known or predicted operon structure for the gene(s) of interest, expression correlation map for the local gene neighbourhood, Objective: An important class of modeling methods in transcriptional polarity across the operon, predicted upstream Systems Biology deals with structural or topological (parameter- intergenic features (terminator, Shine-Dalgarno and/or free) analysis of cellular networks. Hence there is a need for transcription factor binding site, if any) and a compilation of up to appropriate software tools that provide algorithms for analyzing date annotation from the EBI, KEGG, Sanger centre and UniProt both mass-flow as well as signal-flow networks. is given. Results: CellNetAnalyzer (CNA) is a MATLAB toolbox providing Conclusions: The Streptomyces operon prediction server, a comprehensive and user-friendly environment for structural StropE (http://strep-microarray.sbs.surrey.ac.uk/StropE), is analysis of metabolic and signaling networks. The main modeling the first web-tool (for any organism) that allows the viewing paradigm is to represent the networks as hypergraphs. In of predicted/known operons and their expression in terms of the case of metabolic networks the nodes of the graph are polarity (Laing et al, 2006) and respective gene neighbourhood. metabolites and the edges carry reaction fluxes which interconvert The StropE database and interface can be distributed and those metabolites whereas in signaling networks the nodes are implemented for other sequenced prokaryotes. species and the edges represent logical or causal interactions References: between those species. Laing E, Mersinias V, Smith CP, Hubbard SJ. In preparation. CNA provides a large toolbox with various, partially unique, Laing E, Sidhu K, Hubbard SJ. BMC Genomics 2008 9(1):79. algorithms for analyzing mass-flow and signal-flow networks. The Laing E. Phd thesis, University of Manchester 2007. particular strengths of CellNetAnalyzer are methods for functional Laing E, Mersinias V, Smith CP, Hubbard SJ. Genome Biology network analysis, i.e. for characterizing functional states, for 2006 7(6):R46. detecting functional dependencies, for identifying intervention strategies, or for giving qualitative predictions on the effects of perturbations. The main features for metabolic networks are: -computation of graph-theoretical properties -metabolic flux analysis

192 ICSB 2008 DS3-3-56 trichome patterning, growth of epidermis, and growth of an entire leaf. Tools for sharing, analysing and visualising chemical Conclusion: VVE proved to be a useful tool for the modeling of biology data growing tissues, and a step towards the development of a general Persson, Ronnie1; Brive, Lars2 modeling methodology for developing discrete 2-manifolds. 1University of Gothenburg, Department of Chemistry, Göteborg, Sweden; 2University of Gothenburg, Department of Cell- and DS3-3-58 Molecular Biology, Göteborg, Sweden Visualizing the effects of alternative splicing on protein and Objective: The Platform for Chemical Biology (www.cbp.science. domain interaction networks gu.se) aims to characterise molecular function by the regulation of Emig, Dorothea1; Cline, Melissa2; Lengauer, Thomas1; Albrecht, individual signalling pathways using small-molecule compounds. Mario1 We present the communication and analysis tools that were 1Max Planck Institute for Informatics, Computational Biology developed that link data of chemical, biological and 3D structural and Applied Algorithmics, Saarbrücken, Germany; 2University of origin. California, Molecular Cell and Developmental Biology, Santa Cruz, Results: The main parts of the database are microarray data United States (in-house and public), gene information (including, e.g., GOterm functional classification), chemical compounds (e.g., structure, Objective: Recent studies have revealed that alternative splicing physicochemical data, availability) and 3D structure data (crystal plays an important role for protein and interaction diversity. and modelled structures). An important part are the links that Alternatively spliced protein isoforms may vary in their domain relate different types of data, and one can therefore ask questions composition, potentially leading to the loss or gain of protein such as “how are genes in pathway x regulated by compounds interactions. The Affymetrix Exon Array allows for measuring with substructure y?”. Users interact with the data through a web exon expression and thus for studying alternative transcripts interface designed for intuitive use. and their constituent domains. Therefore, the integration of exon Visualisation of sub cellular location of selected gene products expression data with protein and domain interaction networks is Queries of data in the database often result in collections of very valuable for the analysis of the impact of alternative splicing genes that are regulated in a particular way as a result of a stress on protein and domain interactions. event. A new tool was developed that allows the user to map Results: We have developed DomainGraph, a plugin for the these onto a schematic cell with selected organelles. Known free, open-source software platform Cytoscape (http://www. interactions from the BIOGRID database are plotted as lines cytoscape.org) for visualizing and analyzing biological networks. between the gene products. The user can interact with the plot DomainGraph comprises two main functionalities: (1) the Posters to extract information about genes and interactions. Nodes can visual decomposition of a protein interaction network into the Dedicated be colour coded by any type of data, typically M-values from underlying domain-domain interactions and (2) the integration of microarray experiments. Statistical significant functional properties exon expression data for highlighting protein domains and their of included genes can be calculated and plotted. interactions affected by alternative splicing events. This approach Conclusions: The tools described accelerate the work enables the user to easily detect occurrences of alternative process and make it possible to perform complex queries and splicing and to investigate its effects on the protein and domain visualise results in more intuitive ways. They were developed for network. Conclusions: DomainGraph is a powerful tool that Saccharomyces Cerevisiae, but should also be applicable for supports an integrative analysis and visualization of protein and other organisms. domain interactions together with exon expression data. This enables the visual interpretation of alternative splicing events in DS3-3-57 the context of protein and domain networks. Additional analysis methods for the graphical comparison of domain graphs point VVE: A simulation framework for discrete 2-manifold the user to similarities and dissimilarities in expression patterns Barbier de Reuille, Pierre1; Prusinkiewicz, Przemyslaw2 of different tissues or cells. The plugin DomainGraph and the 1INRA, Norwich, United Kingdom; 2University of Calgary, CPSC, comprehensive online documentation and tutorial are available at Calgary, Canada http://domaingraph.bioinf.mpi-inf.mpg.de

Objective: Modeling plant growth is a complex process DS3-3-59 that requires suitable software tools. Their construction and advancement of the underlying theories are still at an early eSOMet: A novel tool for highly reliable biomarker research stage. This poster presents VVE, a modeling software identification that extends and adapts to biological problems vertex-vertex Haddad, Isam1; Frimmersdorf, Eliane2; Benkert, Beatrice1; systems, introduced previously in the context of geometric Schomburg, Dietmar2; Jahn, Dieter1; Hiller, Karsten1 modeling. 1Technical University of Braunschweig, Department of Results: The main data structure of VVE is the graph rotation Microbiology, Braunschweig, Germany; 2Technical University of system: a directed graph, associated with local editing operations, Braunschweig, Department of Bioinformatics and Biochemistry, in which the neighborhood of each vertex is ordered in a cyclic Braunschweig, Germany manner. This structure proved convenient for representing discrete 2-manifolds and locally expressing their development. Objective: Modern high-throughput techniques like GC-MS VVE is implemented as a C++ library and is written using current facilitate the identification and quantification of hundreds of generic programming methods. This makes it possible to reuse metabolites of a biological system, thus covering large parts of existing code in C or C++, and to create modules for all aspects the metabolome. Due to the amount and complexity of obtained of the modeling process. The VVE software includes a number data, there is an increasing demand for the development of of modules, such as a cell-complex library, allowing for easy appropriate computational analysis methods. We present a novel modeling of plant tissues as the cellular level; an embedded ODE analysis pipeline especially designed for high-throughput based solver for numerically solving systems of differential equations metabolomics data which enables the detection of hierarchical (e.g., reaction-diffusion systems) defined on growing discrete relationships within different metabolic patterns measured under manifolds; and a cell-system library for simple specification of various conditions. cell division using grammar-like rules. The VVE software also Results: eSOMet is a new software which implements provides visualization and user interaction facilities, which rely on established algorithms like hierarchical cluster analysis (HCA) the Qt toolkit and the OpenGL library. The user may interactively along with modern methods like emergent self organizing maps manipulate the simulated tissue and attach virtual probes to (ESOM). This makes it highly reliable for the purpose of deducing monitor the state of individual cells. The system was successfully underlying relationships within a series of metabolomics data. The used in a number of applications, which include models of functionality of the tool covers the ability of biomarker discovery,

ICSB 2008 193 detection of statistical outliers and the automatic mapping of DS3-3-61 detected metabolic differences onto KEGG metabolic pathway maps. Bridging the gap between libSBML and MIRIAM: A library In order to validate the described methods we analyzed a for exploiting SBML annotations metabolomics time-series dataset containing in total 126 Swainston, Neil; Mendes, Pedro; Kell, Douglas metabolic patterns of Corynebacterium glutamicum cells grown University of Manchester, Manchester Interdisciplinary Biocentre, on different carbon sources. Although the hierarchical overall Manchester Interdisciplinary Biocentre, 131 Prince, United structure of the metabolic patterns was similarly detected by HCA Kingdom and ESOMs, obvious differences concerning the highly resolved relations were observable and investigated. Objective: Systems Biology Markup Language (SBML) is an Conclusions: The developed analysis pipeline could be established community developed XML format for the markup rendered as a valuable tool for the reliable detection of hidden of biochemical models (Hucka et al., Bioinformatics, 19, 524- structures within GC-MS based metabolome data. Especially the 531, 2003). With the introduction of annotations to SBML level deployment of emergent self organizing maps is an indispensable 2 version 3, developers of such models were given the facility of extension to the spectrum of metabolome data analysis methods. annotating these models such that individual components such eSOMet is freely available as a Java Web Start application at as species or reactions could be assigned with ontological terms. http://esomet.tu-bs.de. These annotations provide the facility to specify definite terms to individual components, allowing software to unambiguously DS3-3-60 identify such components and thus link the models to existing data resources. New possibilities of BioUML workbench LibSBML (Bornstein et al., Bioinformatics, 24, 880-881, 2008) is Kolpakov, Fedor; Tolstykh, Nikita; Lapukhov, Sergey; Kiselev, Ilya; an application programming interface library for the manipulation Shadrin, Aleksey of SBML files. While libSBML provides the facilities for reading Institute of Systems Biology, Design Technological Institute of and writing such annotations from and to models, it is beyond the Digital Techniques, Novosibirsk, Russian Federation scope of libSBML to provide interpretation of these terms. Results: A Java library has been developed to provide the facility Dedicated Objective: reconstruction of complex biological systems requires for interpreting these annotations, acting as a layer on top of Posters integrated software that provides: libSBML that links to appropriate web services that describe • integration with a wide range of biological databases these ontological terms. This library acts as a reusable code • integration with omics data base that can be utilized in numerous tools for performing SBML • powerful search capabilities manipulation. • visual modelling approach Conclusions: The library can be used to annotate unannotated • client-server architecture for team work. models, and an example tool is introduced which exploits the Results: BioUML (http://www.biouml.org) is an open source search facility of ChEBI (Degtyarenko et al., Nucleic Acids Res., integrated Java workbench for systems biology. Its core is a 36, D344–D350, 2008) or UniProt (UniProt Consortium, Nucleic meta model that provides an abstract layer for comprehensive Acids Res., 36, D190-D195, 2008) to assign ontological terms formal description of wide range of biological and other complex to metabolites and proteins respectively. The library can also systems. New version of BioUML workbench provides: be utilized to read annotated models, and a web service is • graphic notation editor - it allows an advanced user to create introduced that extracts synonyms from appropriate web services new graphic notation or extend an existing one. As an example for each metabolite in a model. SBGN graphic notation was created. • composite database module - allows a user to specify set of DS3-3-62 databases that he would like to use simultaneously as a source of components during creation and editing of diagrams. Graphical interpretation and validation tools in multiblock • EBI data pack - database modules for main databases methods for the analysis of complex biological data supported by EBI: Ensembl, UniProt, ChEBI,Reactome, IntAct, Hassani, Sahar1; Martens, Harald1; Borge, Grethe Iren1; Hanafi, GO, BioModels, SBO. These databases are installed on the Mohamed2; Qannari, Mostafa2; Telle-Hansen, Vibeke H.3; Ulven, special BioUML server and can be easily searched and used by Stine Marie3; Norrheim Larsen, Laila4; Kohler, Achim1 BioUML workbench (client). 1Centre for Biospectroscopy and Data Modelling, Matforsk, • Microarray data analysis - allows a user to load his microarray Aas, Norway; 2ENITIAA, NANTES, France; 3Center for controlled data, identify up/down regulated genes and highligt diagram dietary intervention studies, Akershus University College, elements according to microarray data or results of this data Lillestroem, Norway; 4Epigen, Research Centre, Akershus analysis. University Hospital, Oslo, Norway • BioHub - an approach that allows to arrange relationships between different biological objects formally (it is conceptually Objective: Modern instrumentation makes it possible to study similar with MIRIAM standard) and use this information for biological systems from genotype to phenotype, in terms of high- searching. Results of such search can be presented as an dimensional and informative data at different stages along the editable graph. causal chain: DNA, mRNA, proteome, metabolome and other • Composite diagrams - allows a usert to join several existing phenotypes. There is a need to analyze this mass of data to diagrams/models into a bigger one. relate the different data sets to each other in light of background • Experiment diagram type - allows to specify conditions knowledge. Multi-block methods based on so-called latent (initial values, parameters, external stimulus, etc.) for virtual variables can be used for the analysis and integration of several experiments on the diagram. multivariate data blocks (Kohler et al. 2008). Methods based Conclusion: we believe that new version of BioUML workbench on latent variables reduce the variable space into few important will greatly simplify, speed up and improve the process of formal variables and provide an overview over hidden variation patterns. description and reconstruction of biological pathways. The objective of this paper is to apply multi-block methods to a health intervention study and to study different graphical presentation and validation tools in order to reveal correlations between variables and common variation patterns for different data blocks. Results: In the present paper we used so-called block score and global score plots to obtain graphical overview over the main systematic patterns of sample-to-sample variation. For discovering correlations between variables we used correlation loading plots. A validation strategy based on cross-validation/

194 ICSB 2008 jack-knifing has been developed together with graphical DS3-4-10 representation tools, for presenting the random sample and variable variations that occur when segments of samples are Quantifying nonadditivity derived from inter-component taken out in order to test stability of the obtained model. The interactions in the protein translation system proposed methods are illustrated with an example where different Matsuura, Tomoaki1; Kazuta, Yasuaki1; Aita, Takuyo2; Adachi, types of measurements are obtained in a human intervention Jiro1; Yomo, Tetsuya1 study. 1Osaka University, Suita, Japan; 2Saitama University, Saitama, Conclusions: The increasing amounts of data that are obtained Japan by modern instrumentation can easily lead to a mental overflow for the person analysing the data, whereby unexpected but Objective:Living systems are composed of large number of important patterns of co-variation may be overlooked, while components. To optimize the fitness of biological systems e.g.( , other conclusions may represent wishful thinking rather than protein synthesis activity), it is essential to investigate nonadditivity statistically valid inference. The multi-block data modelling derived from the inter-component interactions. Here, we used methods discussed in this paper provide graphical presentation the in vitro translation system as a representative of the biological tools that allow the user to obtain an overview over complex data. system to investigate the properties of nonadditivity, where In addition, graphical tools for statistical validation model are nonadditivity is defined as a discrepancy between the sum of the provided. effects of respective components’ concentration changes on the activity and that caused by the simultaneous changes. Dedicated session 3-4: Model driven Results:As the system used here was prepared by mixing 69 defined components, protein synthesis activity of this system experimental planning can be defined as a “function” of the concentration of 69 components as “variables.” We measured the protein synthesis DS3-4-09 activity for many different combinations of the concentrations of these 69 components, and subjected these data to Bahadur Modeling stochasticity in artificial gene transfer expansion analysis, which converts the experimental data into Youssef, Simon multiple orders of correlations. Through these analyses, we found LMU Munich, Softmatter group, Prof. Rädler, Munich, Germany that larger than two-body inter-component interactions can be approximated to zero. Furthermore, the ratio of nonadditive term Objective: Gene transfection, the introduction of foreign to the additive term was found to be 0.08 on average. We also genetic material into the nucleus of eukaryotic cells, is of demonstrated that these findings could be used to predict the great importance for genetic manipulations in cell culture. data by estimating unknown parameters (Bahadur coefficients) Posters Gene delivery has therapeutic applications in medicine e.g. from a small number of samples, which is useful in designing Dedicated for treatment of previously incurable cancer types and blood an experimental strategy to optimize the activity of the in vitro disorders. Gene transfer has many barriers or steps. A tailored translation system, the principle of which is likely to be applicable design of nanoparticles could optimize the delivery. To this for analysis and optimization of other biological systems. end computational modeling is of great advantage to grasp Conclusions:We presented an efficient strategy to carry out the complexity of the transfer which contains many stochastic an adaptive walk of the fitness landscape in a high dimensional elements. In particular it helps to elucidate the structure function concentration space of a biological system based on the relationship with regard to the crossing of the differing barrier properties of nonadditivity derived from the inter-component types. The goal is to construct a model that describes the interactions constituting the system. experimental gene transfer distribution functions of single cell assays quantitatively. This model will then be used to predict the DS3-4-11 outcome of experiments with differing complexation agents and receptor-specific targeting. Positional preference in the selection of sensory organ Results: We employ quantitative time-lapse fluorescence precursor cells in proneural clusters microscopy to monitor the dynamics of gene expression in single Hsu, Chao-Ping1; Yan, Ching-Cher Sanders1; Chu, M.-T.2; Pi, cell assays, which allow following the response of large numbers Haiwei2 of individual cells in parallel. We evaluate these movies employing 1Academia Sinica, Institute of Chemistry, Taipei, Taiwan; 2Chang- a homemade image analysis software based on ImageJ. This Gung University, Graduate Institute of Basic Medical Sciences, approach reveals the authentic dynamics of single cell gene Taoyuan, Taiwan expression as well as cell-to-cell variability that would otherwise be masked by population-wide measurements. We developed a Objective: The body of an adult fruit fly is covered with evenly mechanistic stochastic model of the gene transfer that includes spaced external sensory organs composed of four cells derived specifically a detailed description of the endosomal transfer in the from a single sensory organ precursor (SOP) cell. SOP formation cytosol and of the poly-A regulation of mRNA degradation. involves expression of proneural proteins encoded by achaete (ac) Conclusions: We can fit each GFP expression time series to the and scute (sc) genes [1]. Initially, all cells in a proneural cluster are same phenomenological function whose parameters vary by a capable of forming a neural precursor, but only one cell actually wide degree. Apparently this process has a global onset behavior forms an SOP. The Notch (N)-Delta (Dl) signaling pathway allows of intrinsic stochastic nature. Furthermore we derive from the for competition among neighboring cells via lateral inhibition [2,3]. distributions of colors in the co-transfection experiment that the While all the pathways identified are able to follow and amplify a arrival of a complex of plasmids is a statistically independent difference among cells, leading to distinct cell fates, it is not clear event described by a Poisson process. We constructed a model where a difference may start from. from this as well as from knowledge from literature that describes Results: Our previous numerical analyses revealed two opposite the experimental outcome for various GFP vectors. driving forces that bias the choice between middle cells and peripheral cells in a proneural cluster [4]. Lateral inhibition favors cells that contact with fewer inhibitory signal-sending neighbors. On the other hand, we found an indirect mechanism that arises from the autonomous regulation of N and the irreversible N-Dl interaction, which decreases the amount of Dl in the peripheral cells and thus suppresses neural fates in these cells. To see how these mechanisms may contribute to a preferential selection of SOPs, we have classified cell positions for the cells in proneural clusters on developing nota. Our preliminary result indicates a preference for an SOP cell in the center regions of a proneural cluster.

ICSB 2008 195 Conclusions: Our numerical analyses have revealed two different of biological processes. The goal of this work was to identify preferences over the position of selecting SOP cells. Experimental networks affected by chemical exposure and track changes in results have revealed a preference in the position of SOP in a these networks as animals recover. cluster. This result indicates that there exist positional factors in Results: We examined brain microarray data from R. norvegicus the SOP formation. Such factors can be further analyzed through treated with 0, 1.2, 12, 24, and 47 mg/kg RDX at different times both computational and experimental means. after exposure (24hr, 48hr, 7d, 14d, 28d and 90d). A correlation [1] S. Artavanis-Tsakonas, M. D. Rand, and R. J. Lake, Science network was derived from a differentially expressed gene list 284, 770 (1999). and was then transformed to compute partial variances and [2] V. Hartenstein and J. W. Posakony, Dev. Biol. 142, 13 (1990). correlations. The undirected model was converted into a partially [3] N. Oellers,M. Dehio, and E. Knust, Mole. Gen. Gene. 244, 465 directed graph by estimating a pair-wise ordering of the nodes (1994). from the data. All edges in the correlation graph with significance [4] C.-P. Hsu, P.-H. Lee, C.-W. Chang and C.-T. Lee, are directed, the direction of the arrow points from the node Bioinformatics, 22, 1375 (2006). with the larger standardized partial variance to the node with the smaller standardized partial variance. Local FDR (0.05) was used DS3-4-12 and partial correlation was estimated for 64 genes. Out of 2016 & 1896 edges or interactions, only 20 & 18 edges were taken Experimental design using artificial measurement data for network generation. The resulting partially causal network Schlatter, Rebekka1; Conzelmann, Holger2; Gilles, Ernst D.2; exhibited a clear ‘hub’ connectivity structure. Some important Sawodny, Oliver1; Sauter, Thomas1 genes in the network (e.g. Msx3, Cacng1, Brs3) are known 1University of Stuttgart, Institute for System Dynamics, Stuttgart, transcription factors which matched with our network model while Germany; 2Max Planck Institute for Dynamics of Complex other important CNS genes (e.g. Ania-7, Hnrpdl, Alad, Gapdhetc.) Technical Systems, Magdeburg, Germany were also hub genes. Conclusion: A credible transcriptional network was Objective: The activation of caspases is a central mechanism recovered from the time series data. The network included in apoptosis. To gain further insights into complex processes like transcription factors and genes with roles in brain function and this, mathematical modelling using ordinary differential equations neurotransmission. Examination of the dynamic changes in Dedicated (ODEs) can be a very powerful research tool. Unfortunately the expression within this network over time provided insight into Posters lack of measurement data is a common problem in building such CNS recovery from traumas. kinetic models, because it practically constrains the identifiability of the model parameters. A simple way to gain insight into a DS3-4-14 mathematical model is to generate any kind of measurement data in silico. We call this procedure artificial data approach and use it Global networks of functional coupling in eukaryotes by to analyse a mathematical model of caspase activation regarding comprehensive data integration different aspects such as identifiability and experimental design. Sonnhammer, Erik; Alexeyenko, Andrey Results: An existing mathematical model of caspase activation Stockholm Bioinformatics Centre, DBB, Stockholm, Sweden during apoptosis was used in order to design future experimental setups that will help to maximise the obtained information. Objectives: To reconstruct complete gene interaction networks in Artificial measurement data with and without measurement error eukaryotes by integrating as much data as possible from different is generated in silico to simulate additional experiments. The types of evidence, and transferring interactions via orthologs. performed analyses reveal that several kinetic parameters are One of the major challenges in this area is to build a model that not at all, or only scarcely identifiable, and that measurements efficiently integrates vast amounts of heterogeneous data. To this of activated caspase 8 will maximally improve the parameter end it was necessary to develop a range of novel methods to estimates. Furthermore, we can show that many assays with adapt data of different types for integration in a Bayesian network. mutant cells only provide redundant information and as such do Results: FunCoup, an optimised Bayesian framework for not have to be carried out. functional coupling discovery, was developed and applied to Conclusions: In comparison with other methods the chosen reconstruct gene interaction networks in eight eukaryotes. Each artificial data approach has great advantages because it can network edge is annotated with confidence scores of representing be used independently of model properties like stability and a specific kind of interaction – physical interaction, protein the information content of all kinds of measurement settings complex member, metabolic or signalling link. The evidence can be analysed easily, including combined information and sources and organismal origins of each edge are also annotated. measurement errors. The approach allows one to make All the networks are downloadable and queryable at http:// conclusions about system properties, identifiability of parameters FunCoup.sbc.su.se. The web server provides detailed graphical and the potential information content of additional measurements and tabular analysis of subnetworks around query genes, as for the used caspase activation model. The latter facilitates well as comparative analysis of orthologous networks in multiple to improve the experimental design of further measurements species. significantly. Conclusions: We present FunCoup 1.0, comprehensive interaction networks for eight eukaryotes: Homo sapiens, M. DS3-4-13 musculus (mouse), R. norvegicus (rat), D. melanogaster (fly), C. elegans (worm), S. cerevisiae (yeast), A. thaliana (Arabidopsis), and Gene networks involved in neurological function inferred Ciona intestinalis (sea squirt). Eight different types of input data using time-series microarray data from rat brain was used: mRNA co-expression, phylogenetic profile similarity, Habib, Tanwir1; Thodima, Venkata1; Zhang, Chaoyang2; Perkins, protein-protein interaction, sub-cellular co-localization, protein Edward3; Deng, Youping1 co-expression, shared transcription factor binding, co-miRNA 1University of Southern Mississippi, Department of Biological regulation by shared miRNA targeting, and domain associations. Sciences, Hattiesburg, United States; 2University of Southern The input datasets came from over 50 sources in the seven first Mississippi, School of Computing, Hattiesburg, United States; organisms, while the Ciona intestinalis network was reconstructed 3US Army Engineer Research and Development Center, entirely via orthology data. As an example, at confidence >0.75 Environmental Laboratory, Vicksburg, United States the human network comprised 122062 links, connecting 8150 genes. Objective: The Central Nervous System is remarkably plastic in its ability to recover from trauma. We examined recovery from 1,3,5-trinitroperhydro-1,3,5-triazine (RDX) induced seizures in Rattus norvegicus through changes in transcriptional networks. Transcriptional networks from time series experiments provide a simple basis for organizing and studying the dynamic behavior

196 ICSB 2008 DS3-4-15 has multiple homologues of the E. coli proteins and hence is significantly more complex. The goal is to develop a systematic Optimizing ligand administration for hematopoietic stem approach for finding the best experiment that will delineate the cell expansion in bioreactor through a population balance network structure. model approach Methodology & Results: To achieve this, we are constructing, Luni, Camilla1; Francis J, Doyle III2; Elvassore, Nicola3 in silico, various possible models of R. sphaeroides chemotaxis 1Università di Padova, Dept. of Chemical Engineering, Padova, based on the current experimental evidence and gene homology. Italy; 2University of California, Santa Barbara, Dept. of Chemical Applying results from optimal control theory, we determined the Engineering, Santa Barbara, CA, United States; 3Dept. of best input (ligand) profile that gives an output which would allow Chemical Engineering, Padova, Italy us to discriminate best between the proposed models and we are testing this input ligand profile experimentally. Model structures Objective: The scarce amount of CD34+ hematopoietic stem that will be unable to reproduce the experimental results will cells (HSCs) in umbilical cord blood (UCB) is restricting their be invalidated, thus closing the loop between modelling and clinical application to pediatric patients. Our goal is to develop a experiment design. We have also developed other discrimination/ computational framework to design rationally the expansion of invalidation methods, based, e.g., on best initial condition HSCs in a bioreactor. In vivo HSCs respond to the environment design or best structural modification design. The final network by binding ligand molecules, present in the extracellular connectivity will be confirmed using biochemical measurements space, to specific receptors, embedded in cell membrane. (SPR / 2 Hybrid etc.,) thus allowing us to validate our approach Formation of receptor-ligand complexes triggers an intracellular using standard methods. signaling cascade that modifies cell behavior. We are exploiting this mechanism to control extrinsically HSC expansion, by DS3-4-17 manipulating the type, amount and timing of ligand administration. The receptor-ligand pair under consideration is c-Kit and Stem Capture of rapid signaling dynamics enabled through a Cell Factor (SCF). modular microfluidic device Results: We developed a mathematical model based on a Rivet, Catherine1; Hirsch, Alison2; Lu, Hang2; Kemp, Melissa3 population balance equation that takes into account the following: 1Georgia Institute of Technology, Electrical Engineering, Atlanta, ligand-receptor reaction, receptor production, receptor-ligand United States; 2Georgia Institute of Technology, Chemical and complex degradation, and cell division. Reaction phenomena Biomolecular Engineering, Atlanta, United States; 3Georgia are described by mass-action laws with parameters derived Institute of Technology, Biomedical Engineering, Atlanta, United from literature, whereas the cell division rate is described by a States mathematical expression deduced from experimental growth Posters curves. The model accounts for the enhancement of cell division Objective: Mathematical modeling of rapid signal transduction Dedicated rate as a function of c-Kit-SCF complex. The simulations events is limited by the quality of experimental data acquired performed describe not only the average properties of the during short timescales. Microfluidic devices offer the ability of cell population, but also the distributions of both receptor and reduced noise due to repeatable sample handling and precise complex number, as functions of time and ligand concentration. mixing and stimulation. To enable accurate quantification of Preliminary experimental results were collected expanding T-cell receptor activation within seconds after receptor ligation, HSCs from human UCB in a stirred bioreactor, in presence of a modular microfluidic chip was designed for controlled rapid flt3/flk2 ligand, thrombopoietin, and interleukin 6, at different mixing, precise stimulation, and lysis. concentrations of SCF. Results: Cells and stimulus are mixed and split into 8 streams Conclusions: Different kinetics of ligand administration result in module 1. The majority of the incubation time occurs in the in a noticeable difference in receptor and complex distributions, tubing leading to module 2, where cells are ruptured with lysis as well as in cell number, which are related to the phenotypic buffer to extract intracellular proteins. With commercial tubings outcome of the cell population. In particular we derive conditions of varying length, a wide range of time points is possible within for reducing cell heterogeneity by modulating ligand addition one device. A staggered herringbone array was used to achieve according to its uptake by cells. The simulation results highlight full mixing of reagents with minimal shear in less than 0.2 sec on the need of developing a proper experimental strategy for the stimulation chip. This design demonstrates repetitiveness of feedback control of ligand concentration in culture. experiments from day to day and device to device with decreased sample handling and high temporal precision. The device yields DS3-4-16 average protein concentrations of 800 micrograms/ml, which is sufficient for off-chip protein analysis with a Bio-plex instrument. Determining signal pathway connectivity using control Jurkat cells flowing through the device without stimulation engineering tools presented a very low level of MAPK activation, showing that Roberts, Mark1; August, Elias2; McSharry, Patrick3; Maini, Philip4; cells are not stressed by the conditions imposed by flow through Armitage, Judith1; Papachristodoulou, Antonis3 the microchannels. With the additional input of antiCD3 (OKT3) 1Universtiy of Oxford, Deparment of Biochemistry, Oxford, United antibody, T cell signaling dynamics of 6 proteins at 8 timepoints Kingdom; 2Universtiy of Oxford, Deparment of Engineering during the first three minutes are quantified in parallel from the Science, Oxford, United Kingdom; 3Universtiy of Oxford, lysates generated by the modular microfluidic device with less Deparment of Engineering Science, Oxford, United Kingdom; than 4 million cells total. 4Universtiy of Oxford, Mathematical Institute, Oxford, United Conclusion: This microfluidic system is widely applicable for Kingdom quantitative assays of protein activation for modeling of signal transduction networks. The modular format allows for a range of Background: Bacterial chemotaxis is the biasing of movement times to be sampled during receptor stimulation. towards regions of higher concentration of beneficial or lower concentration of toxic chemicals. Chemotaxis in E. coli is one DS3-4-18 of the best understood pathways in biology and there is a large amount of experimental data on structures, kinetics, in Novel feedback mechanisms that generate a phase vivo protein concentrations and localisation. However, with an difference in mammalian circadian transcriptional increasing number of sequenced bacterial genomes it becomes oscillatory network evident that the chemotactic sensory mechanism of other Ogawa, Yukino1; Komamura-Kohno, Yuki2; Koike, Nobuya2; Soga, bacteria is much more complex. Tomoyoshi1; Tomita, Masaru1; Tei, Hajime2 Objective: The aim of our research project is to apply results 1Institute for Advanced Biosciences, Keio University, Kanagawa, from control theory to develop novel approaches for designing Japan; 2Mitsubishikagaku Institute of Life Sciences, group of experiments in order to elucidate the biochemical network Chronogenomics, Tokyo, Japan structure of the chemotaxis mechanism in R. sphaeroides, which

ICSB 2008 197 Objective: In mammals, the cellular circadian rhythms are 015 adenovirus infection modulated by CI1040 MEK inhibition. generated on the basis of the transcriptional-translational Validation of this model should generate continuous time state genetic network that consists of several so-called “clock genes” predictions with respect to the timing of CI1040 treatment encoding transcriptional regulators. Of these clock genes, and ONYX-015 infection. Pending successful test of model Clock and Bmal1 encode transcriptional enhancers that bind to predictions, our goal is to elucidate optimization strategies that the “E-box” motif, while Cry1 and Cry2 encode repressors of could offer practical and effective means for minimizing cancer CLOCK-BMAL1 transactivation. Transcriptions of two essential growth. clock genes, Per1 and Per2, are reciprocaly regulated by CLOCK- BMAL1 heterodimers, CRY1, and CRY2 proteins via E-boxes in DS3-4-20 their promoter regions. However, these expression phases are different by approximately four hours. In this study, we aim to Use of structural identifiability analysis to inform investigate generation mechanisms underlying the phase lag by a experimental design for enzyme kinetics combination of in vitro and in silico experiments. Bearup, Daniel1; Evans, Neil2; Dowson, Christopher3; Chappell, Results: Quantitative analyses are essential for the Michael2 comprehensive elucidation of the structure and function of 1University of Warwick, Chemistry, Coventry, United Kingdom; the regulatory network. We examined the mRNA abundances 2University of Warwick, Engineering, Coventry, United Kingdom; of clock genes in cultured cells derived from rat SCN every 3University of Warwick, Biological Sciences, Coventry, United three hours for one day by quantitative real-time PCR method, Kingdom and simulated the dynamics using mathematical models. The classical transcriptional-translational network model could not Objective: The study of enzyme kinetics invariably combines reproduce simultaneously the abundances and oscillatory phases mathematics, chemistry and biology. Experimental techniques of Per1 and Per2 transcripts, and either a negative or a positive allow the concentrations of species in an in vitro reaction to be feedback regulation for Per1 or Per2 transcription, respectively, measured in real time. Numerical fitting algorithms can then be was required. In order to elucidate this regulation in vitro, used to estimate unknown parameters for a given mathematical transcriptional activities of the Per1 and Per2 promoters were model. Structural identifiability analysis of the model is essential to estimated by luciferase reporter assay method. As a result, only discover whether the parameters could be uniquely determined Dedicated Per2 promoter activity was induced by PER1 and PER2 proteins, (or otherwise) from a perfect noise-free form of the data collected. Posters whereas its induction by CLOCK-BMAL1 was approximately By analysing models corresponding to changes of experimental 3 times lower than Per1. A new mathematical model modified technique it is possible to design experiments which produce by the application of analysis results could reproduce the data appropriate for robust numerical fitting. abundances and oscillatory phases of Per1 and Per2 transcripts. Result: Three experimental approaches to studying a simple two Conclusions: It was strongly indicated that the positive feedback substrate enzyme catalysed reaction were considered. The first regulation of Per2 transcription is one of the reasons for the case, using quasi-steady state assumptions and measurement expression phase lag between Per1 and Per2. of reaction product only, proved unidentifiable using a variant of DS3-4-19 the Taylor series approach. Of the four unknown parameters only one was identifiable. The second case, an experiment measuring An experimentally driven quantitative model for predicting reaction product and an intermediate for a complete time series, and optimizing oncolytic adenovirus cancer treatment was found to be identifiable using the Taylor series approach. The Bagheri, Neda1; Shiina, Marisa2; Korn, W. Michael2; Lauffenburger, third case, an alternative experiment for the second case without Douglas A.1 the measurement of an intermediate, proved intractable using the 1Massachusetts Institute of Technology, Biological Engineering, same approach. However use of a novel input/output relationship Cambridge, United States; 2University of California San Francisco, approach showed this model to be identifiable. Gastroenterology and Hematology/Oncology, San Francisco, Conclusions: Typical experimental procedure for enzyme kinetics United States measures steady state concentrations of reaction product. A single time course of such measurements are insufficient for Replication-selective adenoviruses replicate in cells containing numerical fitting to uniquely estimate the unknown parameters certain mutations, motivating their use in targeted gene therapy. for the relatively simple model considered. These results suggest One such adenovirus, ONYX-015, is a promising new therapeutic that a change of experimental technique to measure pre-steady agent for cancer as it preferentially replicates in p53-defective state concentrations is necessary to allow accurate parameter tumor cells. The efficacy of this virus depends on its successful estimation. Additionally a new input/output relationship approach cell entry via the coxsackievirus and adenovirus receptor, CAR. to structural identifiability analysis has been developed which Unfortunately, CAR is down-regulated in highly malignant cells. proves effective in cases where the Taylor series approach is Although pharmaceutical intervention into the Raf-MEK-ERK unable to produce a result. pathway via the CI1040 MEK inhibitor has shown to up-regulate CAR expression and support ONYX-015 entry, it also causes G1 DS3-4-21 cell cycle arrest, preventing the production of new viruses and virus-induced cancer cell death. Reconstructing gene-regulatory networks from time series Objective: Through the experimentally-driven modeling of cancer Geier, Florian1; Timmer, Jens2; Fleck, Christian2 cells subject to CI1040 treatment and ONYX-015 infection, we 1University of Freiburg, Institute of Physics, Freiburg, Germany; aim to characterize and predict observed system dynamics, 2Institute of Physics, University of Freiburg, Freiburg, Germany providing the means to optimize combinatorial cancer treatment. Results: The replication-selective adenovirus cancer therapy Objective: Cellular processes are controlled by gene-regulatory system comprises a set of nonlinear ordinary differential equations networks. Several computational methods are currently used to with both time/state-dependent parameter rate constants and learn the structure of gene-regulatory networks from data. This time/state-independent parameter values. Analysis of time course study focusses on time series gene expression data in order data yields experimentally motivated parameters that define to identify the underlying network structure. We compare the nonlinear system kinetics. For instance, in the colon cancer cell performance of three different network reconstruction methods line HCT116 treated with CI1040, we observe an up-regulation using synthetic data generated from an ensemble of reference of CAR expression at the cell surface which leads to enhanced networks. Data requirements for the reconstruction of gene- adenovirus infectivity. Such experiments motivate the importance regulatory networks are investigated. Additionally, the effect of of infection rate constants that depend on the timing of drug unobserved cellular processes is studied. treatment and virus multiplicity. Results: We identify linear Gaussian dynamic Bayesian networks Conclusions: We have developed a relatively small but and stepwise regression based on F-tests as suitable methods biologically relevant nonlinear differential equation model that for the reconstruction of gene-regulatory networks from time characterizes the dynamics of cancer cells subject to ONYX- series data. Commonly used discrete dynamic Bayesian networks

198 ICSB 2008 perform inferior and this result can be attributed to the inevitable movement of Dictyostelium discoideum by using a single cell information loss by discretization of expression data. Relative tracking system. Using statistical analysis of these traced to the level of observational noise, we give estimates for the data, we characterized the statistical nature with anomalous required amount of gene expression data in order to accurately diffusion in spontaneous cell movement and also input-output reconstruct gene-regulatory networks. Unobserved processes, relationship in electrotaxis under various field strengths. Then like protein-protein interactions, induce dependencies between we applied a generalized Langevin model with non-linear gene expression levels similar to direct transcriptional regulation. decay rate, fluctuation, and memory for the cellular velocity We show that these dependencies cannot be distinguished from dynamics. By numerical simulations of this model in search of transcription factor mediated gene regulation on the basis of gene the parameters corresponding to the experimental data, we expression data alone. successfully reproduced almost all characteristics of spontaneous Conclusion: Currently available data size and data quality cell movement quantitatively. Furthermore, by simply extending make the reconstruction of gene-regulatory networks from gene the model into the one with the bias towards the external field, expression data a challenge. In this study, we compare three we can also reproduce experimental input-output relationship of commonly used reconstruction methods in order to reverse electrotaxis. engineer gene-regulatory networks from time series data and Conclusions : Taking these results together, we revealed the give estimates for their prediction error in dependence of the data consistent relationship between cellular spontaneous movement quality and size. and tactic response, and functional significance of cellular spontaneous activity. DS3-4-22 DS3-4-24 Stoichiometric regulation is a potential point of fragility in cellular systems: A case study in budding yeast cell cycle Digital pattern analysis based on microarray dynamic- Kaizu, Kazunari1; Yoshida, Yuki1; Moriya, Hisao2; Kitano, Hiroaki1 range database for complex developmental model 1The Systems Biology Institute, Cancer Institute, Department of Seita, Jun1; Sahoo, Debashis2; Rossi, Derrick3; Dill, David2; Systems Biology, Tokyo, Japan; 2PRESTO, JST, Saitama, Japan Weissman, Irving1 1Stanford University, Institute for Stem Cell Biology & Regenerative Objective: We previously reported a genetic method named Med, Stanford, United States; 2Stanford University, Computer ‘genetic Tug-Of-War (gTOW)’, which reveals cellular robustness Science & Electrical Engineering, Stanford, United States; against gene over-dosage. gTOW revealed “robustness profile” 3Harvard Medical School, Immune Disease Institute, Boston, of the cell cycle of budding yeast (Moriya et al., 2006), but United States mechanisms behind them remain to be elucidated. In this Posters study, we applied gTOW in gene knockout strains or additional Objective: Current methodology of microarray data analysis is Dedicated gene over-expression (multidimensional gTOW), so as to clarify to compare two or more samples for significantly differentially molecular interactions providing robustness. expressed genes. Unfortunately, this method typically disregards Results: CDC14 (a mitotic protein phosphatase gene) has the the dynamic range of measurements of the genes in question, lowest limit among 30 cell division cycle (CDC) genes. Using a and may thus fail to identify the appropriate genes. The possible computational model (Chen et al., 2004), we predicted that the range of measurements of a gene depends not only on biological fragility of CDC14 was attributed to the unique stoichiometric factors, but also on the particular sequence of the gene, which regulation by the inhibitor NET1. Here we show the fragility was affects the efficiency of hybridization of the mRNA on the array. stabilized by the over-expression of NET1, and interestingly the Even expression levels of gene x are significantly different correlation between their copy numbers showed fine linearity, just between sample A and B, if both expression levels reside lower as predicted in silico. This relationship was confirmed in single half of dynamic range, biological significance would decrease. cell population analysis using flow cytometry. On the other hand, Meanwhile, even expression levels of gene y are not significant ESP1 (a caspase-like protease gene) has much higher limit than between sample A and B, if the expression levels are near its CDC14, though it is considered to be regulated in the same way maximum possible value, gene y might be important for both (namely stoichiometric inhibition by PDS1). In order to reveal sample A and B. To address this issue, we developed Digital components conferring robustness of ESP1, we quantified the Pattern Analysis based on Microarray Dynamic-range Database. upper limit of ESP1 in 20 cdc gene knockout strains. As a result, Results: Estimating the true dynamic range of genes requires we found that the upper limit of ESP1 was remarkably reduced inspecting the measured values on a large set of highly diverse within either cdh1 (an ubiquitin ligase), or clb2 (a mitotic cyclin) arrays. Fortunately, large amounts of array data are now publicly knockouts. In these knockout strains, the predicted linearity available. We have downloaded over 2000 mouse microarray between ESP1 and PDS1 copy numbers was observed. Further datasets (on an Affymetrix platform) and normalized them all analysis using a esp1 mutant indicated that CDH1 and CLB2 together using a standard algorithm to construct a Microarray enhance robustness of ESP1 by two different mechanisms in Dynamic-range Database. Then a threshold separating “low” addition to the stoichiometric inhibition by PDS1. from “high” values was computed using our StepMiner algorithm. Conclusions: Multidimentional gTOW provides insight into Using the dynamic range and threshold computed from this mechanisms conferring or controlling robustness through much larger data set, we have been able to analyze gene molecular interactions and proposes new aspect of modeling. expression from a complex model involving about a dozen cell types from a developmental pathway with multiple branching DS3-4-23 points. Using a significance based on the dynamic ranges of genes, we analyzed microarray data sets of early hematopoietic From spontaneous cell movement to tactic response: development pathways. Search for lymphoid pathway-specific Analysis of electrotaxis in Dictyostelium cells genes successfully reveals known and unknown genes. Takagi, Hiroaki1; Sato, Masayuki J2; Yanagida, Toshio2; Ueda, Conclusion: Digital Pattern Analysis based on Microarray Masahiro2 Dynamic-range Database improves the significance of microarray 1Nara Medical University, Kashihara, Japan; 2Osaka University, data analysis, and expands applicable scope of microarray data Graduate School of Frontier Biosciences, Suita, Japan analysis toward complex biological models.

Objective : Cells can show not only spontaneous movement but also tactic responses to specific environmental signals in relation to their physiological functions. It is important to investigate spontaneous cell movement and its fluctuations quantitatively, and identify the mechanism to rectify it to realize tactic behaviors. Results : For that purpose, we experimentally took a series of cellular trajectories of both spontaneous and electrotactic

ICSB 2008 199 DS3-4-25 Results: We have developed a new generic framework for the analysis of dynamic systems in terms of control coefficients. Optimal experimental design for the identification of Application of the framework shows that crucial moments in dynamical cell subpopulations the cell cycle are not controlled by a single protein or molecular Busetto, Alberto Giovanni; Fischer, Bernd; Buhmann, Joachim process. On the other hand, control is not completely scattered ETH Zurich, Institute of Computational Science, Zurich, over all processes in the network either: Reactions directly Switzerland responsible for the synthesis and breakdown of the delayed response genes (including Cyclin D/Cdk4) as well as the reactions Objective: Experimental observations are often possible only responsible for dephosphorylation of hyperphosphorylated at the population level, despite the discrepancy between singe- retinoblastoma have the highest absolute control on the cells and populations. We define an estimator that permits restriction point. In addition, a strong correlation was observed the identification of heterogeneous subpopulations. Since the between the control of specific variable species and the control measurements are performed only at a finite and scarce number of the restriction point. The reaction steps that control the of time points, being able to optimize the information gain by restriction point most likely do so by affecting the Cyclin E/Cdk2 sampling is highly desirable. Whereas usually in experiments dimer concentration; the restriction point is highly sensitive to the time points for experimental data acquisition are chosen perturbations made to the Cyclin E/Cdk2 dimer concentration, uniformly spaced, we propose a method that optimally selects when perturbed at the restriction point. the measurement times. Based on information theory, this Conclusions: The developed framework based on Metabolic scheme permits the maximum information gain even in the case Control Analysis can be used to analyze the control properties of undersampling. The method finds the set of time points that of cell cycle networks, thereby indicating which processes or minimizes the maximum entropy of the estimated parameter reaction steps exert a high level of control on system properties probability density. The resulting estimation is the most precise such as the restriction point. This framework also aids in between the least biased. understanding the control exerted on high-level system properties Results: The method has been tested on a biologically realistic of the cell cycle in terms of how the system reactions affect function that models the dynamics of protein degradation, the concentrations of model variables (i.e. proteins). This may taking into account enzymatic decay, dilution, transcription and aid experimental design as well as network-based drug target Dedicated translation rates. The computed optimal sampling reduces the prediction. Posters error between the model “ground truth” and the estimate for the probability density and the parametrized initial conditions. OS-03 Entropy maximization avoids an additional bias even in the case of undersampling. Also, the minimization of the entropy between Identification of the gene network governing progression of the least biased results assures the systematic reduction of the atherosclerosis and effects of cholesterol lowering uncertainty associated with data scarcity. The method permits to Lundström, Jesper1; Björkegren, Johan2; Tegner, Jesper2 include prior knowledge that significantly facilitate the optimization 1Computational medicine group, Medicine, Karolinska institute, process. Despite its frequent use, uniform sampling can miss Stockholm, Sweden; 2Computational medicine, Medicine, important dynamical features that the optimized scheme is able Stockholm, Sweden to exploit. Conclusions: Due to its generality, the method can be applied Objective: Atherosclerosis, the most common cause of death to a large number of biologically relevant experimental situations, in the westernized part of the world, is a life long disease where ranging from protein degradation to gene expression. At a small deposits of fatty material develop in the artery walls leading computational cost, the usefulness of the experimental data can to complications such as myocardial infarction and stroke. In be significantly increased. By integrating previous results with a recent study (Skogsberg et al. PLOS Genetics, 2008) we new optimal samplings, we expect that this approach will prove identified more than 1000 genes important for atherosclerosis useful, because it permits a close feedback loop of modeling/ progression. The aim here is to take the next step to identify computation and experimentation. and characterize the gene networks important to progression of atherosclerosis and responding to cholesterol lowering. Other session Results: Literature mining generated a list with 187 seed genes known to be related to atherosclerosis and cholesterol metabolism. We created a network of the seed genes their OS-02 neighborhood (defined as any node connected to a seed gene though a PPI or in the expression network) using three data Subtle control of the restriction point of the mammalian cell sources, protein protein interaction (PPI), literature mining and cycle gene expression data (reverse engineering from expression of Conradie, Riaan1; Bruggeman, Frank2; Ciliberto, Andrea3; Csikász- human aorta biopsies). The resulting network includes 1,948 gene Nagy, Attila4; Novák, Béla5; Westerhoff, Hans6; Snoep, Jacky7 (nodes) interconnected with 25,704 edges. The network was then 1Univeristy of Stellenbosch, Department of Biochemistry, analyzed to identify central modules important to atherosclerosis. Stellenbosch, South Africa; 2Vrije Universiteit, Molecular Cell To facilitate the analysis we also utilize expression data from Physiology, Amsterdam, Netherlands; 3FIRC Institute of Molecular cholesterol lowering in a atherosclerosis mouse model and an Oncology Foundation, Milan, Italy; 4Microsoft Research University association to atherosclerosis. Our analysis revealed both shared of Trento, Centre for Computational and Systems Biology, Trento, and distinct modules related disease progression and cholesterol Italy; 5University of Oxford, Oxford Centre for Integrative Systems lowering respectively. Biology, Oxford, United Kingdom; 6Manchester Interdisciplinary Conclusion: We have performed the first global gene network Biocentre, Manchester Centre for Integrative Systems Biology, analysis which elucidates the important modules governing Manchester, United Kingdom; 7University of Stellenbosch, the progression of atherosclerosis. The network is mostly likely Department of Biochemistry, Stellenbosch, South Africa relevant for developing drugs tailored to hamper the development of atherosclerosis. Furthermore, to the best of our knowledge we Objectives: The cell cycle has been described in many kinetic have also performed the first analysis of the networks responding models, which typically focussed on the protein species rather than to a lowering of cholesterol. Our work on atherosclerosis networks on the processes within these models. In addition, previous studies parallels similar efforts in the fields of other complex diseases focus more on the sequence of cell cycle events than on what such as cancer, diabetes and obesity. control these events. We set out to 1) develop a generic method for the analysis of dynamic systems within the framework of Metabolic Control Analysis and 2) to apply this to a model of the restriction point of the mammalian cell cycle. Our aim was to quantify the importance of the reaction steps on the restriction point.

200 ICSB 2008 OS-04 Both transcriptomics and proteomics analyses were used to gain insight in the effects of metal stress on the molecular level. Combinatorial regulation across species: co-regulatory Conclusions: Preliminary results show a clear impact on all levels associations, hierarchy of regulation and evolutionary of biological organisation. There is a linear correlation between dynamics cadmium exposure and cadmium accumulation in the gills and Bhardwaj, Nitin1; Gerstein, Mark2 significant cadmium accumulation in the liver due to chronic 1Yale University, Molecular Biophysics and Biochemistry, New cadmium exposure while there is a trend to energy depletion in Haven, United States; 2Yale University, New Haven, United States the liver. The condition factor tends to decrease compared to the controls. Furthermore the swimming performance decreases due Relatively small number of TFs can set up strikingly complex to cadmium exposure. Transcriptomics and proteomics analyses spatial and temporal gene expression patterns. Appropriately, using microarrays and 2D-DIGE are currently under way and combinatorial gene regulation by various transcription factors these results will also be discussed. (TFs) cooperatively has been an active area of research. While many context- and subsystem-specific rules have emerged, our OS-06 understanding of the general principles behind combinatorial regulation still remains limited. Here, we build and analyze co- Dynamic network topology changes in functional modules regulation network between TFs for five species displaying a predict responses to oxidative stress in yeast large spectrum of evolution: E coli to Human. We use a network Peddinti, Gopalacharyulu1; Velagapudi, Vidya1; Lindfors, Erno1; transformation procedure to obtain co-regulatory network Halperin, Eran2; Oresic, Matej1 describing associations between TFs to regulate common genes. 1VTT Technical Research Center of Finland, Quantitative Biology Analysis of co-regulatory network reveals many interesting and Bioinformatics, Espoo, Finland; 2International Computer properties, such as the presence of two kinds of regulatory Science Institute, Berkeley, CA, United States hubs: (i) those that make more co-regulatory interactions, thus serving as integrators, and (ii) those that make few co-regulatory Objectives: The topology of metabolic or protein-protein interactions, thereby specifically regulating one or few cellular interaction networks has been extensively studied. Our primary processes. Interestingly, the co-regulatory interactions vary as interest is to investigate how the context, such as changing logarithmic of the number of target genes suggesting that the physiological state of the system, affects connectivity within number of co-regulatory interactions reaches a certain value and cellular functional modules and, through this, obtain better then stabilizes, except for E coli for which the relationship is linear. understating of the control mechanisms of biological systems. We also analyze the co-regulatory network in the context of the Results: We investigated the changes in network connectivity hierarchy of gene regulation and, interestingly, we find that co- as results of oxidative stress. We collected available S. cerevisiae Posters regulatory interactions occur more frequently between the TFs at data on protein-protein interactions (DIP), metabolic pathways Dedicated the same level of hierarchy than between TFs from different levels. (KEGG), gene regulatory relationships (TRANSFAC) and gene Finally, analysis of the evolutionary dynamics of the co-regulatory expression data for oxidative stress response at different time network and estimation the growth of these networks by points. By defining criteria for evaluating presence or absence of duplication, inheritance, loss of homologous interactions and gain proteins, we constructed integrated networks for each time point. of novel ones reveals that these networks grew predominantly We devised methods to identify functional modules using Gene by loss/gain of interactions than by duplication and inheritance. Ontology (GO) membership of protein nodes in the network, and Finally, we find that in spite of huge variation in the size and derived a statistical metric to measure the connectivity changes complexity of these networks in different species, these properties across the time points. We found that ceramide biosynthesis, lipid show relatively lesser variation across species indicating that transport, lipid metabolism are among highly activated modules these qualities might be inherent properties of these networks. during oxidative stress. This activity could not be identified using commonly applied pathway analysis methods such as Gene set OS-05 enrichment analysis. We validated these findings by performing lipidomic analysis of yeast dynamic response under oxidative A systems biology approach to the effects of metal stress stress. in fresh water fish Conclusions: Our results imply that the connectivity changes Benoot, Donald; Vergauwen, Lucia; Timmers, Mieke; Knapen, of the system in response to stress or other external stimuli may Dries; Blust, Ronny determine biological function. University of Antwerp, Biology, Antwerp, Belgium OS-07 Objective: In this ecotoxicological study the biological effects of environmental pollutants are studied using a systems biology Protein evolution is faster outside the cell approach, integrating these effects across different levels of Julenius, Karin1; Pedersen, Anders Gorm2 biological organisation. The first objective of this research is 1Karolinska Institutet, Medical Biochemistry and Biophysics, charting the overall stress response of zebrafish to cadmium. The Stockholm, Sweden; 2Technical University of Denmark, Center for toxic effects of cadmium have been widely studied and include Biological Sequence Analysis, Lyngby, Denmark disturbances of ion and osmoregulation, respiration and enzyme function. A wide range of effects has been described at the gene Objective: The purpose of this study was to analyze mammalian and protein expression level. In most of these studies integration proteins/genes with known subcellular location for variations in of different levels of biological organisation within the same evolution rates. experimental setup was not achieved. Future research focuses Results: We show that proteins that are exported (extracellular on the stress response to metals and organic pollutants via water proteins) evolve faster than proteins that reside inside the cell and food. (intracellular proteins). We find weak, but significant, correlations Results: In our study, cadmium exposure experiments were between evolution rates and expression levels, percentage conducted in the laboratory on adult zebrafish at different of tissues in which the proteins are expressed (expression sublethal cadmium concentrations. The stress response was broadness), and the number of protein interaction partners. More observed at several time points. A selection of parameters was important, we show that the observed difference in evolution rate assessed, over different levels of biological organisation in order between extra- and intracellular proteins is largely independent to create an integrated overview of the stress response, from the of expression levels, expression broadness, and the number of organismal level over the biochemical level to the molecular level. protein-protein interactions. We also find that the difference is not At the highest level, organismal effects were investigated using caused by an overrepresentation of immunological proteins or the critical swimming speed and the relative condition factor. On disulfide bridge-containing proteins among the extracellular data the biochemical level we measured tissue specific energy storage set. We conclude that the subcellular location of a mammalian as well as tissue specific and whole body metal accumulation. protein has a larger effect on its evolution rate than any of the

ICSB 2008 201 other factors studied in this paper, including expression levels/ methods: NIR and ARACNE. This behaviour is seen on all three of patterns. We observe a difference in evolution rates between the source networks provided with SynTReN, using fully knocked- extracellular and intracellular proteins for a yeast data set as well out datasets. and again show that it is completely independent of expression Conclusions: We found a simple new method that is able to levels. accurately reconstruct gene regulatory networks. This method Conclusions: Proteins that are exported evolve faster than performs much better than methods known in literature. proteins that reside inside the cell. This is true both in mammals Furthermore, it is not biased to a certain graph type and much and for yeast species. The difference is not caused by differences faster than the other methods, which makes it possible to analyse in expression and/or number of interaction partners. large datasets. References: OS-08 [1] Bulcke et al., BMC Bioinformatics 2006, 7:43 [2] Gardner et al., Science 2003, 301, 102-105 Sequence similarity and expression similarity in human- [3] Margolin et al., Nature protocols 2006, 1,2,663-672 mouse orthologs Murata, Shinya1; Saito, Rintaro2 OS-10 1Grad. Sch. Media and Governance., Keio Univ., Fujisawa, Japan; 2Dept. Environment & Info. Studies., Keio Univ, Fujisawa, Japan Simulation of fibrillation in a realistic model of canine atrial tissue Objective: Orthologs are defined as pairs of genes derived from Wallman, Mikael1; Jirstrand, Mats1; Jacobson, Ingemar2 the common ancestral gene. Analyses of the relationship between 1Fraunhofer-Chalmers Centre, Göteborg, Sweden; 2Integrative functional, sequence and gene expression similarity of orthologs Pharmacology, AstraZeneca R&D, Mölndal, Sweden are necessary to infer how the ortholog genes diverged from the original genes and how they evolved. Objective: Atrial fibrillation is the most common form of heart Results: In this study, expression similarity of human-mouse arrhythmia and is associated with a five to six fold increase orthologs were analyzed using expression data retrieved from in the incidence of stroke. Computer models describing this SymAtlas. Orthologs between H. sapiens and M. musculus were phenomenon are useful to understand and predict how it is Dedicated identified using RSD method which considers not only global affected by ion-channel blockers. The objective of this ongoing Posters sequence alignment but also evolutionary distance. Homologous project is to investigate the relationship between dynamic gene pairs in human and mouse detected by BLAT, and human- behavior at the tissue level and at the level of individual cells by mouse random gene pairs were used as negative controls (non- means of in silico experiments. orthologs). We found that there were higher expression similarities Results: A framework for modeling and simulation of electro- in ortholog pairs, compared with random pairs. There was chemical activity in large scale cell networks has been developed. positive correlation between expression similarity and statistical Within this framework, a geometric model of the canine atria has significance of sequence similarity (BLAT e-value) in homologous been constructed from ultra sound imaging data. Realistic fiber pairs. However this correlation was not observed in orthologous structure and cell type distribution have also been incorporated pairs. into the model. The simulation framework has been used to Conclusions: Our results suggest that expression patterns (and induce atrial fibrillation like behavior in cell networks and the effect possibly functions) of ortholog gene pairs are conserved even of ion-channel modulation on this behavior has been studied. there are sequence divergences. On the other hand, homologous Conclusions: The modeling approach taken in this work is genes may contain many paralogs and their expression patterns indeed able to qualitatively reproduce the dynamic behavior as well as their functions may have diverged depending on associated with atrial fibrillation. Additionally, this behavior has sequence divergences. been terminated by modulation of certain ion channel parameters, mimicking drug effects. This suggests that the particular OS-09 mechanism of action investigated here is a viable approach for prevention and termination of atrial fibrillation. A new method for reverse engineering gene regulatory networks using a combination of correlation and T-tests OS-11 Boere, Diana W.M.; Bosnacki, Dragan; Ten Eikelder, Huub M.M.; Hilbers, Peter A.J. A comprehensive molecular interaction map of the budding Eindhoven University of Technology, Biomedical Engineering, yeast cell cycle Eindhoven, Netherlands Ghosh, Samik; Moriya, Hisao; Matsuoka, Yukiko; Kaizu, Kazunari; Yoshida, Shimuzu; Kitano, Hiroaki Objective: Reconstructing gene regulatory networks from knock- The Systems Biology Institute, Tokyo, Japan out microarray experiments remains difficult due to the noisy character of the data. Consequently many different methods have Objective: Molecular biologists have painstakingly dissected been developed to solve this problem. Furthermore, due to a and identified the intricate molecular actors and interactions lack of microarray data of a reasonable size of which the original involved in different phases of cell cycle control in the budding network is known, it is hard to test these methods. Our goal is to yeast Saccharomyces cerevisiae. In light of the central role of cell compare various methods on more realistic datasets of hundreds cycle in cell growth and proliferation as well as various disease of genes and to find better methods. pathophysiologies, it is of paramount importance to collate the Results: To generate microarray data of knock-out experiments disparate biological resources into a comprehensive molecular SynTReN[1] was used. We tested some of the most used network which can lend itself to computational and dynamic methods, using the positive predicted value (PPV) and the analyses. sensitivity as criteria. Using their default settings, a method based Results: In this work, we present a comprehensive map of on a paired student T-test came up as the best method, followed the cell cycle for Saccharomyces cerevisiae, capturing the by NIR[2], based on multiple linear regression, and ARACNE[3], molecular entities involved as well as related pathways. The map based on a generalisation of the Pearson correlation. encompasses existing biological knowledge extracted from a However, a new method that consists of a combination of the large number of original papers available on cell cycle control in T-test and the correlation yields up to a two fold increase in yeast. The map is created in CellDesigner software (Version 4.0) the PPV at the same sensitivity compared to a T-test. In this using a well-defined and consistent graphical notation and is method the gene regulatory networks are reverse-engineered stored in the Systems Biology Markup Language (SBML) format. using a linear combination of two P-values. The first P-value was Structural and topological analyses reveal the overall architecture obtained from a paired student T-test on the data. The second of the pathways involved. Moreover, the map is published on was obtained from a T-test on the hypothesis of no correlation. the online, community tagging platform Payao which allows The new method performs much better than the other tested researchers to view and annotate the map in an iterative and

202 ICSB 2008 collaborative manner. highly aggressive tumors causing patient death. The objective of Conclusions: While a large body of work exists elucidating this study was to examine expression profiles in gastrointestinal specific molecular components of cell cycle, we envision that stromal tumors in relation to genetic background and clinical such a comprehensive map for the budding yeast will provide an outcome. integrated, system-level view of eukaryotic cell cycle. It provides Results: We performed gene expression profiling on 17 a platform for systematic study of various molecular pathways GISTs carrying KIT exon 11 deletions or wild type KIT exon and can play a key role in the identification and verification of 11. Microarray screening, using oligonucleotide microarrays drug targets or disease biomarkers involving cell cycle regulation. containing more than 26000 genes covering the entire human Moreover, the availability of the comprehensive map via Payao will genome, resulted in 50 differentially regulated genes between provide a community knowledge base for up-to-date discussions the two tumor groups. Meta analysis of our expression data and exchange of information, thereby accelerating knowledge- compared to data from two public, independent studies of GIST enhancement on eukaryotic cell cycle gene expression (Kang et al. 2005, Subramanian et al. 2004) was performed and confirmed the specific expression profile of GISTs OS-12 carrying KIT exon 11 deletions. Ten of the differentially regulated genes (CD133, DOG-1, KIT, Design principles of gene regulatory networks governing CD34, MMP2, DLK1, CA2, IGFBP5, TGFBI, SERPINF1) were the differentiation processes verified by RT-PCR. CD133 was further analyzed at the protein Kim, Junil1; Kim, Tae-Geon2; Jung, Sung Hoon2; Kim, Jeong- level on a population based tumor material with 204 GIST patients Rae1; Park, Taesung3; Heslop-Harrison, Pat4; Cho, Kwang-Hyun1 from western Sweden using tissue microarray (TMA) and CD133 1Korea Advanced Institute of Science and Technology (KAIST), specific antibodies. CD133 protein was demonstrated in 28 % Department of Bio and Brain Engineering, Daejeon, Republic of all GISTs, notably from gastric location and GISTs showing of Korea; 2Hansung University, Dept. of Information and spindle phenotype. CD133 protein levels were correlated to Communication Engineering, Seoul, Republic of Korea; 3Seoul GIST genotype with KIT exon 11 deletions. Survival analysis National University, Department of Statistics, Seoul, Republic of with Kaplan-Meier curves shows significant correlation between Korea; 4University of Leicester, Department of Biology, Leicester, CD133 protein levels and poor patient prognosis, independent of United Kingdom KIT mutational status. Conclusions: Gastrointestinal stromal tumors having different Objective: Gene regulatory networks governing differentiation mutational status demonstrate specific gene expression profiles. processes are designed to robustly reconstruct multi-cellular A subgroup of GISTs expressed CD133 that correlated to poor organisms from single cells. Although such networks have survival. CD133 may be used as a marker to identify GIST diverse consequences, essential structures of the networks are patients that need extended treatment and follow-up. Posters similar, suggesting evolutionary conservation based on shared Dedicated evolutionary design principles. Our goal is to characterize such OS-14 evolutionary design principles of differentiation control modules. Results: To investigate the design principles, we have The endoderm gene regulatory networks: Subcircuits of computationally evolved random regulatory networks with a Sox32 and Sox17 in zebrafish development preference for hysteresis, or multistationarity, two commonly Yuh, Chiou-Hwa1; Chan, Tzu-Min2; Wang, Horng-Dar2 observed dynamical features of gene regulatory networks related 1National Health Research Institutes, Division of Molecular to differentiation processes. Here, hysteresis means a bistable and Genomic Medicine, Zhunan Town, Miaoli County, Taiwan; switching phenomenon with two switching thresholds depending 2National Tsing-Hua University, College of Life Science and on current state; multistationarity means the characteristics having Institute of Biotechno, Hsin-Chu, Taiwan multiple steady states. We have analyzed the resulting evolved networks and compared their structures and characteristics In vertebrate development, the process of gastrulation leads to with real gene regulatory networks reported from experiments. three germ layers: ectoderm, endoderm, and mesoderm. The We found that the artificially evolved networks have particular endoderm originates from the most marginal blastomeres of topologies and, surprisingly, these topologies share important blastula stage embryos in zebrafish. Nodal signals transduced features and similarities with the real gene regulatory networks, by its receptors lead to the activation of Gata5, Bon and Og9x, particularly in contrasting properties of positive and negative which then activates Sox32 to promote the expression of Sox17 feedback loops. and activates the endoderm differentiation. Here, we established Conclusion: We conclude that the present particular structures the subcircuits of Sox32 and Sox17 using morpholino against of real gene regulatory networks were possibly obtained Sox32 and Sox17, and measured certain gene expression profiles through evolution to achieve multistationarity or hysteresis. by quantitative real time RT-PCR and validation by using in-situ To enforce hysteresis, the networks might have been evolved hybridization. We identified several interesting functional motifs to decrease the number of NFLs while maintaining PFLs. To that are important building blocks in zebrafish developmental enforce multistationarity, the networks might have been evolved Gene Regulatory Networks (GRNs). At early stage, Sox32 and to decrease the number of NFLs and to increase the number of Sox17 autoregulatory lock on the endoderm fate. Early activation PFLs. turns to late repression for Sox32 to Sox32 itself and Otx2 to Sox17. Late transcription factors repress early activator by Gata5 OS-13 activates Sox17 and then Sox17 represses Gata5; Gbx1 activates Sox32 and then Sox32 represses Gbx1. We also found Sox32 Differential expression of CD133 in gastrointestinal stromal and Sox17 repress many early transcription factors such as tumors (GIST) is related to KIT mutational status and FoxH1, and Sox17 represses itself at later stage. patient survival Furthermore, we uncovered the relevant Sox17 cis-regulatory Arne, Gabriella1; Kristiansson, Erik2; Nerman, Olle3; Nilsson, elements, and examined the specific input predictions of the Bengt4; Ahlman, Håkan4; Nilsson, Ola1 GRNs. We discovered three conserved modules: A, B, and C, 1Sahlgrenska University Hospital, Pathology, Lundberg Laboratory with a synergistic effect among them. It was revealed that Pou5f1 for Cancer Research, Göteborg, Sweden; 2University of binding element on B module and Sox32 binding element on C Gothenburg, Zoology, Göteborg, Sweden; 3Chalmers University module works synergistically. An evolutionary non-conserved of Technology, Mathematical Statistics, Göteborg, Sweden; R module exhibits a repressive effect on both the ventral and 4Sahlgrenska University Hospital, Surgery, Göteborg, Sweden dorsal side. Together, we demonstrated directly the structural and functional relationships of the genomic code at this key node of Objective: Gastrointestinal stromal tumor (GIST) is the most the endoderm GRNs in zebrafish development. This information common mesenchymal tumor of the gastrointestinal tract and provides a new insight to the complexity of endoderm formation is caused by mutations in the KIT or PDGFRA genes. GISTs and serves as a valuable resource for the establishment of a represent a spectrum of clinical diseases from benign tumors to complete endoderm gene regulatory network.

ICSB 2008 203 OS-16 on supported lipid membrane structures. The first example that will be covered is the exchange of lipid material between Statistical estimation of errors in gene expression data charged supported lipid membranes and liposomes; potentially a arising in course of confocal scanning method for the in situ modification of supported membranes. In Myasnikova, Ekaterina1; Surkova, Svetlana1; Gursky, Vitaly2; particular, it was possible to follow the different steps during such Samsonova, Maria1; Reinitz, John3 modification processes both by QCM-D and TOF-SIMS, the latter 1St. Petersburg State Polytechnical University, St. Petersburg, allowing direct estimation of the fraction of different lipids in the Russian Federation; 2Ioffe Physico-Technical Institute, St. membrane. The second example involves the use of supported Petersburg, Russian Federation; 3Stony Brook University, Stony lipid membranes to investigate how nanoparticles interact with Brook, NY, United States membranes, as a way for rational design of new nanocarriers for biomolecular drugs. The third example covers how lipases act on Objective: The confocal scanning microscopy in conjunction with membranes, and in particular how lag phases for such interaction fluorescent labeling is a powerful tool for acquisition of accurate can be monitored. Depending on the type of lipase under study, and standardized quantitative data on gene expression at a either dissolution or remodelling of the membrane was observed. resolution of a single cell. However the use of this technique for Conclusion: The combination of surface-supported lipid systems biology is limited due to possible experimental errors. membranes and surface-sensitive analytical techniques allows The aim of this work is to estimate the most essential errors which for detailed studies of processes of relevance for biological arise in the course of fluorescence quantification and to set the membranes. In particular, the molecular composition can be range of experimental conditions allowing to obtain the sufficiently controlled, and morphological changes of the membrane accurate data. structure can be induced and visualized. Results: In photon-limited confocal imaging the major source of errors is Poisson noise due to the discrete nature of photon OS-18 detection. The common way to reduce this noise is the averaging of multiple frames. We have developed a method for the Gene expression in complex tissues: Validating in-silico extraction, estimation and removal of the residual photon noise microdissection from the averaged images. Averaging of images may also lead Repsilber, Dirk1; Kern, Sabine2; Selbig, Joachim3; Jacobsen, Dedicated to the biased data in case of clipped single frames. A method Marc4 Posters based on censoring technique is used to estimate and correct 1Research Institute for Farm Animal Biology, Genetics and this kind of errors. One more source of errors lies in the image Biometry / Biomathematics group, Dummerstorf, Germany; segmentation procedure which results in the underestimation of 2University of Potsdam, Potsdam, Germany; 3Inst. Biochemistry expression levels when is applied to blurred confocal images. We and Biology, University of Potsdam, Bioinformatics, Potsdam, have developed a deconvolution method which allows to estimate Germany; 4Bernhard-Nocht Institute for tropical Medicine, and correct these errors and to make the quantification procedure Hamburg, Germany less sensitive to the accuracy of image segmentation. Conclusions: An important application of the current work is the Objective: Gene expression experiments in multicellular possibility to corroborate the high accuracy of a large amount of organisms, e.g. humans, higher plants etc., involve isolating quantitative gene expression data accumulated in our previous transcripts from tissues. Tissues are a mixture of different cell study and stored in the FlyEx database (http://urchin.spbcas. types. Hence, analysing gene expression data from tissues ru/flyex) as well as to give recommendations how to achieve comes with problems concerning interpretation and confounding the required accuracy of the data in the context of the further with cell type proportions. For some tissues and not too large work. It has been proved that the choice of proper experimental sample sizes microdissection is a possible solution, such that conditions provide the high data quality sufficient to use the only homogeneous parts of the tissue in question are analysed. dataset to study mechanisms of pattern formation, to infer However, this comes with additional costs and is not possible at regulatory interactions in the genetic network and to develop new all for most tissues. mathematical models. Under certain circumstances cell type specific gene expression This work is supported by NIH grant RR07801, GAP award patterns and cell type proportions can be estimated from a RUB1-1578, NWO-RFBR project 047.011.2004.013, and RFBR decomposition of the original data matrix. Approaches along this grants 08-01-00315a, 08-04-00712a. line already exist, but do not consider realistic values for noise and differential gene expression, are not honestly validated with OS-17 experimental data (on both tissue and isolated constituting cells) and do not consider sample size estimation for studying gene Supported lipid structures as biomimetic model systems expression in complex tissues. Svedhem, Sofia1; Kunze, Angelika2; Ekstrand, Helena2; Sjövall, Results: We have improved existing approaches with respect Peter3; Frost, Rickard2; Edvardsson, Malin2; Kasemo, Bengt2 to these issues and also concerning important algorithmic 1Chalmers University fo Technology, Applied Physics, Göteborg, details. Both simulation study and application to an experimental Sweden; 2Chalmers Universty of Technology, Dept of Applied validation dataset on gene expression in blood PBMC of Physics, Göteborg, Sweden; 3SP Technical Research Institute of different groups of tuberculosis patients and their contacts will Sweden, Borås, Sweden be presented. From samples sizes as low as 20 per group a deconfounding approach seems applicable. Especially those Objective: Engineering of surface-supported lipid membrane genes which are regulated in more than one cell type, and in model systems is currently a very active field of research. The different directions, are found differentially epressed with much present contribution will present a number of recent examples higher power. from our group in this area, including both different kinds of Conclusions: Gene expression studies with systems biological supported lipid structures; both (planar) supported lipid bilayers ambition have to rely either on single cell organisms or have to and (intact) supported liposomes; as well as different kinds of face the problem of heterogeneous tissues when studying OMICS biomolecular interactions associated with them. Key experimental profiles. Our example for gene expression in blood demonstrates techniques used to follow these processes are the quartz crystal the feasability of an in-silico microdissection approach for gene microbalance with dissipation monitoring (QCM-D), surface expression in blood. We can show that, given typical sample sizes plasmon resonance (SPR), fluorescence microscopy, atomic for clinical studies or surveys, interpretation and analysis of gene force microscopy (AFM) and time-of-flight secondary ion mass expression data from heterogeneous tissues can be enhanced spectrometry (TOF-SIMS). or at least supported by using a decomposition approach as Results: We will focus on three cases: (i) Lipid exchange between proposed. liposomes and supported lipid membranes of opposite charge, (ii) interactions between nanoparticles designed for drug delivery and biomimetic supported membranes, and (iii) the action of lipases

204 ICSB 2008 OS-20 OS-22

Small RNAs can operate in large complex networks Spectral karyotyping analysis (SKY) of rat endometrial MacLean, Dan1; Elina, Nataliya2; Heimstaedt, Susi1; Havecker, adenocarcinoma Ericka3; Studholme, David4; Baulcombe, David3 Falck, Eva1; Nordlander, Carola2; Behboudi, Afrouz2; Klinga Levan, 1The Sainsbury Laboratory, Bioinformatics, Norwich, United Karin1 Kingdom; 2MRC Clinical Sciences Centre, London, United 1Department of Biomedicin, School of Life Science, University Kingdom; 3Cambridge University, Plant Sciences, Cambridge, of Skövde, Skövde, Sweden; 2Dept of Medical Genetics,, SU/ United Kingdom; 4The Sainsbury Laboratory, Bioinformatics, Sahlgrenska, Gothenburg, Sweden Norwich, United Kingdom Objective: Endometrial adenocarcinoma (EAC) is a disease with Objectives: Small RNAs regulate gene expression by causing serious impact on the human population. The inbred BDII rat degradation of mRNA or inhibition of translation. sRNAs are strain represents a good model for analysis of EAC in humans. produced from double-stranded RNA precursors by Dicer BDII female rats are highly prone to develop EAC, more than 90% enzymes. The sRNAs target other RNAs by sequence and of BDII virgin females spontaneously develop EAC during their life allow cleavage by AGO class nucleases. The resulting cleavage span. products go onto prime further RNA degradation1. In this study Spectral karyotyping (SKY) is a powerful technique that paints we examine the idea that there are multiple cascades of sRNAs all the chromosomes in different colors and allows them to be in a large network. We generate a network of Arabidopsis sRNAs visualized in a single experiment. In this technique, chromosome by predicting targets of sRNA populations obtained by high- rearrangements, such as numerical and structural aberrations, throughput sequencing experiments. can easily be detected so that small translocations can be seen Results: The topology of the network shows it to have a as a transition from one color to another at the chromosomal power law connectivity distribution 2 , to be dissortative 3, breakpoint region. In the SKY technique metaphase preparations highly clustered 4 and composed of more components than an of tumor cells or cell lines are used for the spectral analysis. In equivalent random network. Examination of sRNA populations the present study, a set of 23 rat endometrial adenocarcinoma in the Arabidopsis tex1 sRNA mutant 5, shows that the network was selected for cytogenetic analyses using SKY technique. exists functionally in vivo. Kmer repetition within the genome The aim of the study was to find particular chromosomes that network structure is inherited from and made possible by the were recurrently involved in different forms of chromosomal underlying repetitiveness in the genome sequence. abnormalities. The results are of fundamental significance and Conclusions: Together our results provide good evidence for pave the way for continuation of the project towards finding the existence of a large, robust small RNA interaction network and defining candidate chromosomal regions and genes with Posters with distinct regulatory function. Such a network could have a important role in EAC development. Dedicated massive effect on the regulation of gene expression via mediation Results: The data obtained from analysis show specific and of transcript levels and could be a major contributor to cellular recurrent patterns of chromosomal changes among subgroups information processing and organismal complexity. of tumors. We detected recurrent translocations involving rat chromosome 10 (RNO10) RNO3, RNO6 and RNO1. Interestingly, References a number of these translocation events were commonly observed 1 Baulcombe, D.C. RNA Silencing in plants. Nature 431,356-363 at specific cytogenetic bands of the chromosomes involved. (2004) Conclusions: These interesting findings suggest that some of 2 Barabási, A.L., Albert, R. Emergence of scaling in random the translocations may have been directed toward production of networks. Science 286,509-512, (1999). oncogenic activation of specific cancer-related genes located at 3 Redner, S. Networks: teasing out the missing links. Nature 453, the breakpoints. 47-48 (2008) 4 Watts, D.J.; Strogatz, S.H. Collective dynamics of ‘small-world’ OS-23 networks. Nature 393, 409–10 (1998) 5 Elina et al, Characterisation of the tex1 mutant. In Preparation Systematic Analysis of Genetic Interaction of Escherichia coli OS-21 Mori, Hirotada1; Wanner, Barry2; Takeuchi, Rikiya1; Yamamoto, Natsuko1; Nakahigashi, Kenji3; Yanagihara, Fusamitsu1 Analysis of relative mutation frequency in Escherichia coli 1Nara Institute of Science and Technology, Graduate School using a novel mutability index of Biological Sciences, Ikoma, Japan; 2Purdue University, Kim, Hyunchul; Lee, Baek-Seok; Tomita, Masaru; Kanai, Akio Department of Biological Sciences, West Lafayette, United States; Institute of Advanced Biosciencs, Tsuruoka city, Japan 3Institute of Advanced Biosciences, Keio University, Tsuruoka, Japan Mutations are essential for the evolution of organisms and known to be happening spontaneously. If spontaneous mutations Robustness is an important fundamental property of biological preferentially avoid potentially essential amino acid residues in systems. Common mechanisms that give rise to robustness each protein, it will confer advantage to survive to the organisms. depend on alternative or bypass pathways of metabolism In order to evaluate the correlation between mutability and and other redundant biological processes. Also, elucidation of essentiality, we first developed a nucleotide level mutability epistatic relationships among genes can give insight into the index based on the previous study (Wright et al. (2004)) and understanding of physiological networks. Synthetic lethality further expanded to the amino acid residue level mutability index or sickness by double gene knockout mutations is one of the considering the non-synonymous codon usage in each amino most powerful methods for analyzing robustness. To make this acid residues. We applied the index to Escherichia coli proteome possible in Eshcrichia coli, which is one of the best organisms to and found the potentially essential parts such as conserved understand cellular systems comprehensively based on the vast residues, buried residues, essential proteins, and highly amount of accumulation of biological knowledge, we setup an expressed genes, which have slow evolutionary rates, had lower easy and reliable system for construction double knockout strains mutability than their counterparts. These results indicated that by conjugation and for analysis of their growth effects. selection force act on amino acid residue mutability associated We previously established the comprehensive single gene with essentiality and suggest that organisms have been increasing knockout library, called Keio collection. To combine the chances for survival by reducing mutagenesis on potentially second deletion mutation with a single gene deletion of Keio essential parts even before certain selections. collection, we initiated construction of a second single gene deletion library carrying a different (chloramphenicol) resistance cassette with additional features, including (1) turbo GFP fusion with the initiation codon of the target gene, (2) a modified FLP

ICSB 2008 205 recombination site (FRT1) site, does not recombine with FRT, OS-25 and (3) a 20-nt molecular bar code downstream of the targeted gene. We are also developing tools for efficient conjugation for Simple molecular network designs leading to optimal combining different single deletions to make double knockout responses under oscillatory stimulations mutants. To do this, a fragment carrying tra genes andoriT of Cournac, Axel; Sepulchre, Jacques-Alexandre wild-type F plasmid, which are essential for conjugation and University of Nice-Sophia-Antipolis, Institut non Linéaire de Nice, transfer, were combined with oriRγ replicon. This plasmid can Valbonne, France replicate in cells supplying pir gene product. The chromosomal fragment of the target site of integration of modified F, is cloned Objective: Bacteria or cells receive a lot of signals from their by homologous recombination. The resultant pseudo F plasmid environment and from other organisms. In order to process this is transferred by conjugation to the newly established single gene massive information, Systems Biology show that a central role is deletion mutant and selected Hfr strain integrated F plasmid played by regulatory networks composed of genes and proteins. by antibiotic resistance selection. In these ways, we are now We aim at presenting and discussing simple regulatory network developing the high-throughput system of conjugation. We motifs having the property to maximize their responses under will report the present situation and show preliminary results of time-periodic stimulations. analysis of genetic interaction. Results: We consider basic network motifs studied in the literature, as the interlerlocked negative feedback loop and the OS-24 incoherent feedforward foop (IFFL) (U. Alon 2006). Using standard mathematics and numerical simulations, we explain the types of Metabolic engineering of Saccharomyces cerevisiae for responses that the IFFL can exhibit with respect to activation by biochemical production from xylose a train of periodic pulses. We show that this system has a non- Scalcinati, Gionata1; Otero, Josè M.1; Van Vleet, Jennifer R. H.2; vanishing response only when the inter-pulse interval is above Jeffries, Thomas W.3; Nielsen, Jens1; Olsson, Lisbeth1 a threshold. A slight generalisation of the IFFL (the diamond) 1Chalmers University of Technology, Department of Chemical is shown to work as an ideal passband filter. In these systems and Biological Engineering, Goteborg, Sweden; 2University of the average response in the concentration of the output protein Wisconsin-Madison, Department of Bacteriology, Madison, United presents a maximum for a specific pulse pattern. This prediction Dedicated States; 3Institute for Microbial and Biochemical Technology, Forest is illustrated with actual biological data, and comparison is Posters Products Laboratory, Wisconsin, United States made between prokaryotes and eukaryotes. We next study the interlerlocked negative feedback loop, i.e. a 2-gene motif forming Objective: Interest in the use of renewable raw materials like a loop where the nodes respectively activate and repress each lignocellulose has risen dramatically in hopes of developing a other. We analyse the situations where this system displays a bioprocess production platform competitive with fossil fuels. resonance phenomenon under periodic stimulations. Finally, Saccharomyces cerevisiae presents many advantages for combining the previous results we show a mechanism by which lignocellulose bioconversion, including fast sugar consumption the response amplitude in a molecular network can be maximized rates and resistance to inhibitor compounds often found by using a bursting temporal pattern. in lignocellulose hydrolysates. Lignocellulosic materials are Conclusions: We present several simple designs of molecular composed of a mixture of hexoses and pentose sugars. Hexoses, networks producing optimal output in response to periodic such as glucose, are readily metabolized, but pentoses, such as stimulations of the system. The identified mechanisms are simple arabinose and xylose, are not natively consumed by S. cerevisiae and based on known network motifs in the literature, so that as primary substrates. Xylose utilization is of commercial that they could be embodied in existing organisms, or easily interest for efficient conversion of lignocellulosic materials into implementable by means of synthetic biology. Moreover we bioproducts; however, S. cerevisiae is not able to consume show that these designs can be studied in different contexts xylose at industrially relevant rates. The aim of this study was the of molecular biology, as for example in genetic networks or in construction of an efficient xylose consumingS. cerevisiae cell signaling pathways. factory for bio-based chemical production coupled to biomass formation. OS-26 Results: In this study S. cerevisiae has been metabolically engineered to utilize xylose through overexpression of the Multicriteria optimization in systems biology: Application to xylose catabolic pathway from Pichia stipitis. After prolonged, metabolic engineering and flux balance analysis repeated batch shake flask cultivations on xylose, evolvedS. Sendín, José-Oscar H.; Alonso, Antonio A.; Banga, Julio R. cerevisiae with improved xylose uptake kinetics was obtained. I.I.M. (CSIC), Process Engineering Group, Vigo, Spain The selected strain grew in chemically defined medium under aerobic conditions on xylose with a fast specific growth rate, Objective: Mathematical optimization plays a key role in high biomass yield and no xylitol accumulation. The strain also computational systems biology. In the majority of applications, exhibited improved xylose consumption in glucose-xylose however, only a single performance index is considered, but real mixtures, with glucose remaining the preferred substrate. world optimization problems are inherently multiobjective: two or Conclusions: The resulting metabolically engineered S. more criteria, often in conflict, must be satisfied simultaneously, cerevisiae strain exhibits one of the fastest growth rates on xylose and optimal trade-offs have to be found. This work is aimed at reported in the literature for strains where xylose reductase, showing the benefits of multiobjective (or multicriteria) optimization xylose kinase, and xylose dehydrogenase were over-expressed. (MOO) in biological applications by solving two different case Furthermore, the resulting xylose consuming strain is a promising studies: i) metabolic engineering of the amino-acid biosynthesis microbial cell factory for chemical-biomass coupled production in bacteria (involving modification of regulation and enzyme given that 64% of all carbon is diverted to biomass. Systems expression levels); and ii) multiobjective flux balance analysis for biology approaches, widely applied to S. cerevisiae, are required predicting intracellular fluxes inEscherichia coli. to develop mechanism based understanding and obtain further Results: Pareto fronts (i.e. the set of optimal trade-offs between improvements. the objectives) are obtained for both case studies using a novel MOO method which extends and improves classical techniques and can be easily linked to well-known SBML tools. For the metabolic engineering problem, where the aromatic amino-acid production and enzyme and metabolite levels are optimized, this approach not only provides optimal alternatives for genetic manipulation, but also it can be used to detect the most important regulatory loops. For the FBA problem, simultaneous optimization of growth, ATP yield and enzyme usage (without any other constraint) produces an optimal set of solutions which are

206 ICSB 2008 able to predict flux distributions for both aerobic and anaerobic data. The most popular strategy is using internal reference cultures in different environmental conditions. genes. However, finding appropriate reference genes for data Conclusions: In this contribution, a MOO approach is used normalization is one of the most challenging problems today. to solve two different case studies for which the solutions Results: In the present study, the effect of the different describing the Pareto front improve the results obtained when the pre-processing strategies on results of cluster analysis was optimization of a single cellular objective subject to constraints investigated. For this purpose dynamic expression profiles of is considered. One of the advantages of this approach is that 22 genes involved in glucose sensing, signal transduction and no additional, case-specific, and often artificial constraints are glucose repression pathways were used. The data was obtained required. We show that this strategy is also a powerful tool by RT-PCR as a response to system level perturbations in 4 for a better understanding of the factors that influence the different S. cerevisiae BY4743 strains. It was seen that choice metabolic flux and is able to cope with the complexity and lack of of pre-processing method could have an important effect on the knowledge in models from SB. results. Work supported by EU project BaSysBio LSGH-CT-2006-037469 Conclusion: The genes that fall into same cluster differ according to whether the data were normalized with respect to the steady OS-27 state expression level or auto-scaled or just normalized with respect to a reference gene. EWOD-based digital microfluidics as a novel platform for systems biology OS-32 Shin, Yong-Jun; Lee, Jeong-Bong University of Texas at Dallas, Electrical Engineering, Richardson, COSMIC - Systems biology of Clostridium acetobutylicum - United States A possible answer to dwindling crude oil reserves Millat, Thomas1; COSMIC Consortium2 Objective: Droplet-based or digital microfluidics involves precise 1University of Rostock, Department of Computer Science, manipulation of discrete and independently controllable droplets, Rostock, Germany; 2The Netherlands, United Kingdom, Germany and it enables us to perform biological experiments on a chip in micro scale. This paper reports a novel integrative way of Clostridium acetobutylicum is a commercially valuable bacterium, studying enzyme kinetics, one of key building blocks of systems first isolated from corn in 1912 by Chaim Weizmann. It is a Gram- biology, using EWOD (ElectroWetting-On-Dielectric)-based positive, sporulating, obligate anaerobe capable of producing digital microfluidics as a novel platform. We used a colorimetric acetone, butanol (biobutanol), and ethanol (A.B.E fermentation) enzymatic reaction-based glucose assay as an example and from a variety of carbon sources (e.g. starch). The metabolism demonstrated high-throughput imaging study of reaction kinetics of C. acetobutylicum is characterized by the Acetone-Butanol- Posters on digital microfluidic devices. Computer simulation of the kinetics Ethanol (ABE) fermentation. Exponentially growing cells mainly Dedicated was carried out, and the result was compared with that of the produce the organic acids acetate and butyrate. During the experiment. We also demonstrated a novel way of controlling transition phase C. acetobutylicum switches towards the droplet motion using machine vision. It was successfully applied generation of the solvents butanol and acetone as dominant to shuttling of the droplet, one of mixing operations of digital fermentation products, a process called solventogenic shift. The microfluidics. results of the experiments show that the crucial prerequisite for Results: The two droplets containing glucose, enzymes and the induction of this metabolic change is the external pH. The other reagents were merged on a digital microfluidic device. definite inducement of the switch is not figured out completely One of the final products was oxidized o-dianisidine which has yet. brown color. Real-time high-throughput imaging was done using The aims of this project will be progressed through a combination Vision Assistant (National Instruments, USA) to quantify color of disciplines - genetics, transcriptomics, proteomics, change over time. The first image was saved in a buffer, and metabolomics, biochemistry, chemical engineering and following images were subtracted by the first image to visualize mathematical modeling - deployed by a consortium of eleven the change only. Red color plane was then extracted, and the European scientists, from the Netherlands, the United Kingdom, mean value was calculated. Total number of images processed and Germany. for the duration of 600 seconds was 18,190 (approximately 30 The overall goal is to establish C. acetobutylicum as the images/second). The normalized experiment result matched to paradigm for clostridial systems biology. The focus lies on key the result of simulation carried out using MATLAB SimBiology. We regulatory and metabolic events that occur during the transition have also demonstrated machine vision-based droplet shuttling, between acidogenic, vegetative growth and the onset of solvent an operation that can accelerate chemical reaction rate. A droplet production and sporulation. Specific workpackages seek was driven back and forth repeatedly by circular edge detection to identify and quantify the changes taking place and which of the droplet. regulatory mechanisms are involved, determine the role of cell Conclusions: This paper reports EWOD-based digital density in the transition, evaluate the significance of redox state microfluidics as a novel platform for systems biology research, and glycosylation and establish how the organisms reacts to which was utilized to demonstrate high-throughput imaging of stress. The goal for mathematical modeling is to unravel the the reaction kinetics and machine vision-based droplet motion regulatory mechanisms leading to the changes in the metabolism control. of C. acetobutylicum. Starting from known metabolic pathway maps, we establish mathematical models that explain the pH- OS-30 dependence of the pathway elements.

Analysis of dynamic expression profiles using pre- OS-33 processed quantitative PCR data Cankorur, Ayca1; Kirdar, Betul2 Functional organization of the yeast proteome by a novel 1Bogazici University, Department of Chemical Engineering, yeast interactome map Istanbul, Turkey; 2Bogazici University, Istanbul, Turkey Valente, André X. C. N.1; Roberts, Seth2; Gao, Yuan2; Buck, Gregory2 Objective: Expression profiling by real-time PCR is a widely used 1Biocant and University of Coimbra, Unidade de Sistemas method since it has many important advantages to expression Biológicos, Cantanhede, Portugal; 2Virginia Commonwealth profiling with microarrays. In terms of data quality and sensitivity University, Center for the Study of Biological Complexity, real-time PCR is superior to microarrays. However, an accurate Richmond, United States method of normalization of real-time PCR data is very important to control the errors in quantization of the changes in mRNA level Objective: It is hoped that comprehensive mapping of protein which are associated with the nature of real-time PCR analysis. physical interactions will facilitate novel insights, regarding There are several proposed strategies for normalizing RT-PCR both fundamental cell biology processes and the pathology of

ICSB 2008 207 diseases. To fulfill this, finding good solutions to two issues will potassium channel openers are known to have cytoprotective role prove essential: i) how to obtain reliable interaction data in a high- but their vascular effect is not well documented.

throughput setting and ii) how to structure interaction data in a Results: We tested NS1619 BKCa opener on calcium meaningful form, amenable and valuable for further biological homeostasis in EA.hy 926 cells and found that NS1619 caused research. an increase in Ca2+ level in a concentration-dependent manner. Results: In this work, we structure an interactome in terms of Moreover, NS1619 decelerated the rate of respiration in EA.hy predicted permanent protein complexes and predicted transient, 926 endothelial cells as measured with oxygen electrode and non-generic, interactions between these complexes. The induced a drop in mitochondrial membrane potential as measured interactome is generated via an associated novel computational with the use of JC-1 fluorescent indicator. algorithm, from raw high-throughput affinity purification/mass Conclusions: In summary, the potassium channels opener spectrometric interaction data. NS1619 displays endothelial action that is characterized by an Conclusions: We apply our technique to the construction of increase in the intracellular calcium level as well as by deceleration a new interactome for S cerevisiae, showing it yields reliability of respiration and a fall in mitochondrial membrane potential.

typical of low-throughput experiments, out of high-throughput Contribution of mitochondrial BKCa channels to all these effects

data. We discuss relevant biological questions in the context of still remained to be confirmed. Nevertheless, BKCa seem to this novel interactome including, via homology, how it relates to represent an attractive target for endothelial drugs that could be human disease. exploited therapeutically.

OS-36 OS-38

Development of high-throughput microphotography system Carrier mediated uptake of drugs: A model system in yeast using microfluidic chip for large-scale phenotyping in Lanthaler, Karin1; Dobson, Paul D.2; Oliver, Stephen G.3; Kell, budding yeast Douglas B.2 Ohnuki, Shinsuke; Nogami, Satoru; Ohya, Yoshikazu 1University of Manchester, School of Chemistry, Manchester, University of Tokyo, Integrated Biosciences, Chiba, Japan United Kingdom; 2University of Manchester, School of Chemistry, Manchester, United Kingdom; 3University of Cambridge, Dedicated Large-scale phenotyping based on cell morphology is one of Department of Biochemistry, Cambridge, United Kingdom Posters powerful methods to reveal relationships between gene function and phenotypic traits because the cell morphology reflects Objectives: Current models of xenobiotic uptake across various intracellular process that depend on gene function. In biological membranes are dominated by lipid diffusion, with budding yeast, high-dimensional and quantitative morphological examples of transporter-mediated uptake often thought to be data can be obtained from microphotographs by using the image exceptional. We describe several lines of evidence that suggest processing program gCalMorph h we developed. Quantitative this view is overly simplistic and which point towards transporters morphological analysis for all of haploid non-essential gene being the major route of xenobiotic uptake. mutants identifies many morphological mutants and relationships Results: The evidence includes hundreds of examples from the between gene functions and cell morphology (Ohya et al., 2005). literature of transporter-mediated uptake of drugs, toxins and From the results of replicated experiments of 58 calcium-sensitive other substrates that appear ideally suited to lipid diffusion and mutants in medium with or without calcium, we assumed that yet are shown to utilise and often absolutely require transporters classification of large-scale mutants toward prediction of gene to cross cell membranes. We screened wild type S. cerevisiae functions is feasible if we obtain replicated morphological data of against more than 4000 drugs in a high-throughput growth all mutants under various conditions (Ohnuki et al., 2007). assay and assessed the concomitant reduction of maximum Although numerous experiments are required to the classification specific growth rate. Cytotoxic compounds (IC50 ≤50μM) were to predict gene functions, the throughput of our current system subsequently screened at their IC90 concentration against the is relatively low, because all steps including cell culture, staining, homozygous diploid deletion pool of S. cerevisiae [1]. Knockout and microscopic observation are done by hand. To improve the strains that exhibit enhanced resistance to the drug out-compete throughput of microphotography system, a late-limiting step, we more sensitive strains (assays of strain ratios performed using chose microfluidics as a core technology because conventional Affymetrix TAG4 DNA arrays [2]). Where transporter knockout high-throughput methods were unsuitable for our image analysis strains exhibit enhanced resistance to the drug relative to wild because of low magnification of images (Rammohanet al., type the transporter is implicated in drug activity. This high- 2006). We designed the microfluidic chip that acquires the throughput assay forms the basis for further detailed analysis of images with four cycles: flow, catch, capture, and release. Using potential drug-transporter interactions. this chip, we able to acquire cell images within two minutes par Conclusions: Our own experiments in yeast have clearly strain (fifteen-fold faster than by hand). They were successfully identified drug-carrier relationships. Based on evidence we analyzed by CalMorph, obtaining consistent data with that of our gathered from the literature and experimental evidence we manual method. In addition, we improved CalMorph software conclude, that carrier mediated drug uptake is rather the rule than that enables to process phase-contrast images and to analyze the exception. data from live cell. Morphological transition during cell cycle 1. Winzeler, E.A., et al., Functional characterization of the S. could be monitored using the developed microfluidics system cerevisiae genome by gene deletion and parallel analysis. and software. In summary, our microfluidic system, together with Science, 1999. 285(5429): p. 901-6. the CalMorph software, will enable us quantitative morphological 2. Pierce, S.E., et al., A unique and universal molecular barcode phenotyping, as well as monitoring of live cell morphology. array. Nat Methods, 2006. 3(8): p. 601-3.

OS-37 OS-39

Effects of NS1619, large conductance potassium channel Is GSM1 regulated by HAP complex? opener, on endothelial cells Dereli, Elif1; Dikicioglu, Duygu2; Kirdar, Betul2 Lojek, Agnieszka1; Wrzosek, Antoni1; Dolowy, Krzysztof2; 1Bogazici University, Chemical Engineering Department, Istanbul, Szewczyk, Adam1 Turkey; 2Bogazici University, Istanbul, Turkey 1Nencki Institute of Experimental Biology, Warsaw, Poland; 2University of Life Science SGGW, Warsaw, Poland Objective: HAP4 is the subunit of the transcriptional regulator HAP complex which was identified as upregulating the expression Objective: The aim of this study was to determine endothelial of cytochrome C and several genes encoding TCA cycle and

effect of large conductance potassium channel (BKCa) opener, respiratory chain enzymes. GSM1 is a yeast gene with unknown NS1619, on endothelial cell line EA.hy 926. Healthy endothelium function whose product is a transcription factor from the zinc plays crucial role in various aspects of vascular homeostasis and cluster family. GSM1 includes a HAP complex binding site in

208 ICSB 2008 its promoter and a cytochrome C heme-binding signature in its OS-42 protein sequence. GSM1 is a member of energy metabolism group, particularly the respiratory chain subgroup, and involved Variations in clique and community patterns in protein in the regulation of energy metabolism, specifically oxidative structures during allosteric communication: Investigation phosphorylation. The objective of this study is to investigate the of dynamically equilibrated structures of methionyl tRNA transcriptional responses of HAP4, GSM1, MIG1 and MIG2 genes synthetase complexes to transient perturbations under glucose and nitrogen limitations Ghosh, Amit; Vishveshwara, Saraswathi via real time rt-qPCR by cultivating three homozygous deletion Indian Institute of Science, Molecular Biophysics Unit, Bangalore, strains of S. cerevisiae (hoΔ/hoΔ, mig1Δ/mig1Δ and mig2Δ/ India mig2Δ). Results: In this study the expression profiles ofHAP4 , GSM1, Objective: The allosteric concept has played a key role in MIG1 and MIG2 were analyzed. Under glucose and nitrogen understanding the biological functions of proteins. The rigidity/ limitations, GSM1 exhibited very similar expression profile to plasticity and the conformational population are the two important HAP4 in hoΔ/hoΔ. HAP4 was activated just after glucose pulse ideas invoked in explaining the allosteric effect. Although molecular injection and GSM1 was activated later in mig1Δ/mig1Δ. In insights have been gained from a large number of structures, a mig2Δ/mig2Δ grown under glucose limitation, GSM1 and HAP4 precise assessment of the ligand induced conformational changes shared the same expression profile before and after the pulse in proteins at different levels, ranging from gross topology to injection. intricate details, remains a challenge. In the present study, we Conclusions: The investigation of expression levels of the have explored the conformational changes in the complexes of mentioned genes provided further evidence that GSM1 is Methionyl tRNA synthetase (MetRS) through novel parameters regulated by HAP complex. Similar expression levels of GSM1 such as cliques and communities, which identify the rigid regions and HAP4 both in hoΔ/hoΔ and mig2Δ/mig2Δ throughout the in the protein structure networks (PSNs) constructed from the non- glucose limited cultivations indicated a possible role of GSM1 in covalent interactions of amino acid side chains. MetRS belongs to regulation of respiratory pathway of yeast together with HAP4. In the aminoacyl tRNA Synthetases (aaRSs) family that play a crucial mig1Δ/mig1Δ, up-regulation of HAP4 (and GSM1) with nitrogen role in the translation of genetic code. pulse suggested a possible role of HAP4 (and GSM1) during Results: These enzymes are modular with distinct domains on the shift from nitrogen limited conditions to nitrogen excess which extensive genetic, kinetic and structural data are available, conditions. highlighting the role of inter-domain communication. The network parameters evaluated here on the conformational ensembles OS-40 of MetRS complexes, generated from molecular dynamics

simulations, have enabled us to understand the inter-domain Posters Evolutionary conservation of human drug targets in communication in detail. Additionally, the characterization of Dedicated organisms used for environmental risk assessments conformational changes in terms of cliques/communities has Jauhiainen, Alexandra1; Gunnarsson, Lina2; Kristiansson, Erik3; also become possible, which had eluded conventional analyses. Olle, Nerman1; Larsson, D.G. Joakim2 Furthermore, we find that most of the residues participating in 1Chalmers University of Technology, Mathematical Statistics, clique/communities are strikingly different from those that take Göteborg, Sweden; 2The Sahlgrenska Academy at the University part in long-range communication. of Gothenburg, Department of Neuroscience and Physiology, Conclusion: The cliques/communities evaluated here for the Göteborg, Sweden; 3University of Gothenburg, Department of first time on PSNs have beautifully captured the local geometries Zoology, Göteborg, Sweden in their detail within the framework of global topology. Here the allosteric effect is revealed at the residue level by identifying the Objective: Pharmaceuticals are typically found in very low important residues specific for structural rigidity and functional concentrations in the aquatic environment. Accordingly, flexibility in MetRS. This ought to enhance our understanding of environmental effects clearly assigned to residual drugs are the functioning of aaRS in general. consistent with high affinity interactions with conserved targets in affected wildlife species rather than with a general toxic OS-43 effect. Thus, evolutionarily well-conserved targets in a given species are associated with an increased risk. We propose MOSBIO: A modular education approach for systems that orthology data of human drug targets can add important biology information in order to direct future research efforts to assess the Stalidzans, Egils1; Mozga, Ivars1; Roos, Christophe2; Fernandez, ecotoxicological risks posed by pharmaceuticals. Elia3; Font, Julio3; Vanags, Juris4; Remm, Maido5; Wolkenhauer, Results: In this study orthologs for 1318 human drug targets Olaf6; Schmitz, Ulf6 were predicted in 16 species of which several are relevant for 1Latvia University of Agriculture, Biosystems Group, Jelgava, ecotoxicity testing. The conservation of different functional Latvia; 2Medicel Oy, Espoo, Finland; 3NorayBio, Derio/Vizcaya, categories of targets was also analyzed. Zebrafish had orthologs Spain; 4Association of Biotechnology of Latvia, Riga, Latvia; to 86% of the drug targets while only 61% were conserved in 5University of Tartu, Department of Bioinformatics, Tartu, Estonia; Daphnia and 35% in green alga. The predicted presence and 6University of Rostock, Department of Computer Science, absence of orthologs agrees well with published experimental Rostock, Germany data on the potential for specific drug target interaction in various species. Objective: Development and application of systems biology as Conclusions: Based on the conservation of targets we propose a interdisciplinary field is limited by the lack of multisdisciplinary that aquatic environmental risk assessments for human drugs specialists. MOSBIO (MOdular education for interdisciplinary should always include comprehensive studies on aquatic Systems BIOlogy) project aims to help the educational process vertebrates. Furthermore, individual targets, especially enzymes, of monodiscipline (biology, chemistry, physics, mathematics, are well conserved suggesting that tests on evolutionarily distant information technology) to become aware of the interdisciplinary organisms would be highly relevant for certain drugs. We propose needs in systems biology. that the results can guide environmental risk assessments Results: MOSBIO is funded by the European Union Lifelong by improving the possibilities to identify species sensitive to learning program Leonardo da Vinci as a Transfer of Innovation certain types of pharmaceuticals or to other contaminants that project. The project involves six partners from Estonia (University act through well defined mechanisms of action. Moreover, we of Tartu), Finland (Medicel Oy), Germany (University of Rostock), suggest that the results can be used to interpret the relevance of Latvia (Latvian Biotechnology Association and Latvia University existing ecotoxicity data. of Agriculture, coordinator) as well as Spain (NorayBio). Three partners are universities, two partners are enterprises (SME) and one partner is a professional association. The Project started in November 2007 and ends in October 2009. Its website is www.

ICSB 2008 209 mosbio.eu Results: RS was added to mBMMC that were prepared by To facilitate the entry into the large area covered by systems culture of BM cells in RPMI supplemented with fetal bovine serum biology using distance learning methods, different topics are (FBS) in the presence of mouse recombinant IL3 (rmIL3) for 8 wks presented in clearly structured educational modules, enabling and then transferred to diverse culture condition prepared by the the suggestion of an individual sequence of training modules. combination of culture media (RPMI, IMDM), animal serum (FBS, Navigation software will recommended a sequence of modules horse serum (HS)) and cytokine mixtures (rmIL3 and recombinant taking into account existing expertise of the particular specialist as mouse stem cell factor (rmSCF)). Cell growth was analyzed by well as the chosen target competence. Thereby, different target CCK-8. Cell differentiation was observed by measuring levels groups (research and development specialists, technologists, of mouse mast cell protease I (MMCP1), TNFα, IFNγ and IL4 in scientists, teachers and managers) can choose modules covering cell lysate and cell culture supernatant. mBMMC growth was the same topic on different complexity levels, corresponding to more efficient in IMDM supplemented with FBS (FSIMDM) or the respecitve existing competence. HS (HSIMDM) than in RPMI supplemented, in the presence of Conclusion: The EU project “Modular Education for rmIL-3. However, the growth was decreased by 0.75 time with Interdisciplinary Systems Biology” with six partners from five the addition of RS to FSIMDM (RFSIMDM), or by 0.78 time with countries is devoted to the development of structured modular RHSIMDM. In contrast, MMCP1 production was elevated by educational materials with a navigation tool allowing individually 3.4 times with RFSIMDM and by 5.4 times with RHSIMDM, in suitable educational steps for a monodisciplinary specialist to comparison with media supplemented with FBS or HS. The raise the competence in systems biology. most efficient productions of TNFα, IFNγ and IL4 were observed in FSIMDM among the analyzed culture conditions. These OS-44 productions were decreased by 0.77 time in TNFα, by 0.62 time in IFNγ, and by 0.5 time in IL4 with RFSIMDM. In contrast, the Design and implementation of a high-throughput RNAi productions were increased by 5.7 time in TNFα with RHSIMDM. screen for endocytosis in drosophila SR+ cells Conclusions: These results suggested that RS have a factor Dey, Gautam; Gupta, Gagan; Gowrishankar, Swetha; Thattai, to control mBMMC differentiation, which is neither rmIL-3 nor Mukund; Mayor, Satyajit rmSCF. National Centre for Biological Sciences, Cellular Organization and Dedicated Signaling, Bangalore, India OS-46 Posters

Objective: It is increasingly evident that endocytosis in metazoan Characterization of the zoarces viviparus liver cells occurs through multiple specialized yet interconnected transcriptome using massively parallel pyrosequencing pathways. For many of these though, in particular clathrin Kristiansson, Erik1; Asker, Noomi1; Lars, Förlin1; Larsson, D.G. independent (CI) modes of endocytosis, little is known about Joakim2 their regulatory circuits and molecular machinery. We have 1University of Gothenburg, Department of Zoology, Göteborg, used the adherent Drosophila SR+ cell line and high-throughput Sweden; 2University of Gothenburg, 2Department of fluorescence microscopy to screen endocytic phenotypes Neuroscience and Physiology, Göteborg, Sweden using RNAi to 7216 conserved genes. We have simultaneously monitored two endocytic pathways, the canonical clathrin mode Objective: The teleost Zoarces viviparus (eelpout) lives along marked by endocytosis of transferrin and the CI GEEC pathway the coast in the cold waters of Northern Europe. The many marked by fluorescent dextran. desirable characters of this fish species, including its stationary Results: Cells were grown and assayed on a customized glass behaviour and giving birth to live young, make it well-suited slide followed by semi-automated image acquisition. Custom- for ecotoxicological testing and environmental monitoring. written MATLAB routines were used to identify single cells and However, the complete lack of genome sequence data renders extract quantitative measures of endocytic uptake. The design whole-genome assays, such as gene expression microarrays, of this screen provides quantitative data at both the single cell impossible. and population level. Heterogeneous responses to RNAi within a Results: We have used massively parallel pyrosequencing to single population of cells allowed us to develop novel statistical characterize the liver transcriptome of Z. viviparus. One single indicators of perturbation. These indicators are robust to the run on a 454 Life Science FLX sequencer resulted in more than systematic noise characteristic of such large-scale experiments. 350.000 reads with an average length of 220 base pairs. A Conclusions: Candidate genes from the primary screen are tailored pipeline, specially optimized for quality control, assembly being channeled through independent confirmatory experiments and annotation of short sequence reads, where developed. which have served to validate both the set of ‘hits’ and the The analysis resulted in 77.4 million base pairs of high quality statistical methodology. The analysis of quantitative phenotypic transcriptome data assembled into 36.000 contigs and 17.000 data from cells and their populations in large-scale screens may singlets, comprising more than 50.000 putative eelpout be generally useful for exploring biological networks. transcripts. Of these, ~ 60% could be annotated against known proteins and the vast majority could be aligned to at least one OS-45 of the five available fish genomes. Interestingly, a considerable Changes of mouse bone marrow derived mast cell amount was located in regions without any previous annotation differentiation by rat serum in mouse mast cell protease I, indicating the presence of novel undiscovered fish genes. Finally, TNFα, IFNγ and IL4 we also present a newly developed 15k eelpout oligonucleotide Lee, Ju Hyun1; Shin, Kyung Sook2; Kim, Hyun Sook2; Guk, Sang microarray, specifically aimed for identification of gene expression Mi2; Joo, Kyung-Whan2; Lee, Joon-Sang2; Kim, Chul Whan3; Cho, profiles in ecotoxicological research. Sung-Weon2 Conclusions: Characterization of a eukaryotic transcriptome 1Korea University Graduate School, Microbiology and without any genomic data a priori is a notoriously difficult Immunology, Seoul, Republic of Korea; 2Korea University College undertaking. We show that one single run of a modern of Medicine, Parasitology, Seoul, Republic of Korea; 3Korea pyrosequencer generates enough high quality data for an University College of Medicine, Pathology, Seoul, Republic of adequate de novo assembly of transcripts. We also show that Korea the resulting sequence information is rich enough for producing Objective: Mast cells are the important cells to contribute to a high-density oligonucleotide microarray for large-scale gene immune reaction control and infection defense, which are derived expression assays. from bone marrow (BM). To enhance the comprehension of mast cell development, the effect of rat serum (RS) on mouse BM derived mast cell (mBMMC) was investigated, as the animal cell culture with species specific serum was regarded to have a similar environment to in vivo and RS were regarded as alternatives to mouse serum.

210 ICSB 2008 OS-48 Conclusions: Our structured mining approach results in the identification of genetic networks relevant in homing. Widespread conservation of genetic redundancy across a billion years of eukaryotic evolution OS-50 Vavouri, Tanya; Semple, Jennifer; Lehner, Ben Centre for Genomic Regulation, EMBL/CRG System Biology Unit, Study of experimentally observable state variables of Barcelona, Spain biological systems through joint analysis of biological and technical hypothesis of the time-lapse tissue culture Objective: Genetic redundancy means that two genes perform development experiment the same function, a situation that arises following gene Stys, D.; Urban, J.; Levitner, T.; Vanek, J.; Timr, S.; Machlica, L.; duplication. Intuitively this redundancy should not be stably Marsalek, J.; Khrytankova, I.; Brezina, V. maintained in genomes, because either gene duplicate can lose University of South Bohemia, Institute of Physical Biology, Nove a shared function without any phenotypic effect. However this Hrady, Czech Republic assumption has never been systematically tested. Here, we comprehensively analyse the origins of genetic redundancy in two The time-lapse microscopy experiment is among the best species from two kingdoms of life. representations of biological cell dynamics. It is currently generally Results: We estimate a conservative date of duplication for 275 accepted that cell fate is best described by the chaotic attractor pairs of genetically redundant genes from yeast and C. elegans and that the observable cell states represent individual basins . Surprisingly we find that nearly all known examples of genetic of attraction. It is, however, not quite clear in which relation the redundancy have been stably maintained for more than a hundred experimentally observable macroscopic parameters are to the million years of evolution. Remarkably, in multiple examples state variables of the attractor space. In other worlds it may be redundancy has been conserved since the divergence of the said that the experiment cuts the state space by a fraction of animal, fungi and plant kingdoms over 1-billion years ago, and space, possibly of lower dimensionality and of unknown shape we show that this conservation normally occurs for the same with respect to system state variables. In practice this means that duplicates in multiple eukaryotic lineages. proper description of biological systems is impossible without Conclusions: We conclude that whilst genetic interactions parallel examination of experiment technical hypothesis. between unrelated genes have been found to evolve rapidly The realistic approach towards objective analysis of the between species, genetic interactions between duplicated genes experiment is to determine evolution of information fluxes can be conserved over extensive evolutionary periods. We assessed by the experiment in time. We present here two suggest that this paradoxical conservation of redundancy can be essential representations of the system: (1) Image of information reconciled by a piggyback mechanism where redundant functions content of each timeframe represented by in values of information Posters are co-selected with non-redundant functions. We conclude that entropy and (2) map of information content fluxes. The former Dedicated genetic redundancy is not only a transient consequence of gene map (1) is assumed to be in (unknown) relation to representation duplication, but is often an evolutionarily stable state. of a point in multidimensional state/phase space of our dynamic biological system, the later map (2) to be in relation to image of OS-49 local Lyapunov exponents or eigenvalues of Jacobian matrix in actual point in state/phase space. Such detailed analysis is highly Mining literature and gene expression databases for novel computationally intensive, however, it might be of high value genes relevant in homing for rapid diagnostic in medicine, biotechnology and any other Strömberg, Anna; Gustafsson, Thomas; Montelius, Andreas discipline utilizing cell biology results. On the way of theoretical Karolinska Institute, Dept. of Laboratory Medicine,Clinical assessment, algorithm development and testing, practically useful Physioly, Stockholm, Sweden tools have been developed such as development experiment setup and diagnostic tools, semi-automated identification of key Objectives: Our goal is to find genetic networks, including cell regions, image comparison tools etc. novel genes relevant in homing, possibly targets for therapy. To Intermediate results of the analysis applied to selected common achieve this goal we have devised a structured approach that cell lines / tissue cultures are presented on the poster. can be applied to a broad variety of biological questions where mechanisms can be described in terms of genetic networks and transcriptional control. Results: Tissue remodelling and regeneration after injury or ischemia is dependent on the recruitment of cells from the circulation into the peripheral tissue - a process termed tissue homing. Homing of cells occurs in a step-wise manner and consists of adhesion, rolling, migration into the tissue and differentiation. A dysfunctional tissue homing of cells with stem cell characteristics from the blood stream has been suggested to correlate with risk factors for several diseases such as coronary artery disease, and thus it is crucial to understand the regulatory machinery underlying homing. Our strategy is to define biological processes that are part of homing. For these sub-processes we collect genes and genetic networks with proven relevance. The resulting genetic network is expanded by structured literature and database searches. We use the network to query large number of gene expression datasets (GDS) deposited in the NCBI GEO database. We use a software tool, Genetic Network Explorer (GNE) to visualize gene expression data superimposed on genetic networks. Our definition of genetic network is in this case restricted to genes affecting the expression level of genes. Literature data on such genetic networks include interactions confirmed by several independent experiments as well as chromatin immunoprecipitation experiments and predicted transcriptionfactor binding sites. Transcriptional regulation as well as gene ontology features on network and individual gene level is assessed for several GDS, resulting in a refined network by exclusion of genes.

ICSB 2008 211

Author Index Index Author Please note that * indicates the presenting author Ates, Ozlem DS2-4-18* Aubrey, Wayne DS3-4-08 A August, Elias DS2-2-80*, DS3-4-16 Avva, Jayant DS2-1-38 Abdali, Salim DS2-3-13 Aymerich, Stephane A-30 Abi-Gerges, Najah DS1-3-27 Aziz, Zeinab DS1-1-28, DS2-4-06 Abrighach, Hicham DS1-1-12 Abu El Einin, Hanaa DS1-1-54, DS3-1-15 B Ackermann, Joerg DS2-2-125 Adachi, Jiro DS3-4-10 Babu, Mohan DS3-1-08 Adiels, Martin DS3-3-43 Bachmann, Julie DS2-1-26, DS2-1-33* Adler, Annette A-07 Badmi, Kalyan Venkatesh DS2-2-123 Aebersold, Ruedi A-39, DS1-1-38 Bae, Jae-Bum DS2-3-16 Ahlman, Håkan OS-13 Baek, Songjoon DS2-2-22 Ahmadpour, Doryaneh DS2-1-111 Bagh, Sangram DS3-1-11, DS3-2-13* Aisaki, K.I DS2-1-34 Bagheri, Neda DS3-4-19* Aita, Takuyo DS3-4-10 Bajic, Vlad DS2-3-07 Akhurst, Tim A-27, DS3-3-35* Bakker, Barbara A-06, DS1-1-63, DS1-3-01, Akira, Shizuo DS2-1-102, DS2-1-85 DS1-3-08*, DS2-4-14, Albeck, John DS2-1-51, DS3-1-09, DS2-1-93, DS1-1-61 DS3-4-03 Baldi, Pierre DS2-2-66 Alber, Mark DS3-3-37 Balestrieri, Chiara DS2-1-63 Alberghina, Lilia A-34*, DS2-1-59*, DS2-1-63, Balsa-Canto, Eva DS2-2-05*, DS3-3-48* DS2-1-66, DS2-1-71, Baltanás, Rodrigo DS3-1-02 DS2-2-95, DS2-2-96 Balwierz, Piotr DS2-1-124* Albermann, Christoph DS2-4-09 Banchereau, Jacques DS2-1-31 Albert, Marie-Astrid DS1-3-08 Banga, Julio R. A-30, DS2-2-05, DS3-3-28, Albertsson-Wikland, Kerstin DS2-3-10 OS-26 Albrecht, Mario DS1-2-08, DS3-3-58 Bao, Xiaoming DS1-1-60 Alcarraz-Vizán, Gema DS2-3-21 Bar, Nadav S. DS2-1-84* Aldridge, Bree DS3-4-03* Barberis, Matteo DS2-2-130, DS2-2-95*, Alexander, Ross A-03* DS2-2-96*, DS2-2-97*

Author Alexe, Roxana DS2-4-33 Barbier de Reuille, Pierre DS3-3-57* Index Alexeyenko, Andrey DS3-4-14 Bardwell, Lee DS2-1-75* Alexopoulos, Leonidas DS2-1-06, DS2-1-60* Barillot, Emmanuel DS2-1-57 Alexopoulos, Leonidas G DS2-2-69 Barisic, Sandra DS2-2-90 Alfieri, Roberta DS2-2-96 Barker, Nathan A-45, DS3-3-02 Alice, Villeger A-51 Barrass, David A-03 Alirezaie, Javad DS1-4-32 Barreiro, Esther A-54 Alkhamesi, Nawar DS2-3-09 Bar-Shira, Anat DS2-3-05 Almeida, Camila DS2-4-23* Barton, Richard DS2-2-133*, DS2-3-09 Almquist, Joachim DS2-1-103, DS2-2-08*, Bastmeyer, Martin DS2-2-37 DS2-2-63* Bauer, Alexander DS2-1-23 Alonso, Antonio A. DS2-2-05, OS-26 Baulcombe, David OS-20 Altafini, Claudio DS1-1-41*, DS2-2-53, Baumann, Ute DS2-1-26 DS2-3-11* Baumeister, Wolfgang P-02* Amano, Akira DS3-3-30 Bayani, Nora DS2-1-70 Ammerer, Gustav DS2-1-03 Bearup, Daniel DS3-4-20* Ananiadou, Sophia A-09*, A-13, A-43*, A-51*, Beaudouin, Joel DS3-4-04 DS3-3-52 Becker, Holger DS2-2-63 Andersen, Ann Zahle DS1-1-51 Becker, Verena DS2-1-26* Andersen, Melvin E. DS2-2-128 Beggs, Jean A-03 Anderson, Alexander DS1-1-33 Begley, Kimberly DS3-3-50 Andersson, Björn DS2-3-10 Behboudi, Afrouz OS-22 Andréasson, T. DS1-4-21 Beisiegel, Ulrike DS2-2-50 Andreu, Lorena DS3-1-17 Beites, Crestina L. DS3-1-16 Andrews, Brenda DS2-4-03* Bekkal Brikci, Fadia DS2-2-105* Anguissola, Sergio DS2-1-108 Belic, Ales DS1-1-44 Anné, Jozef A-22* Beltran, Gemma DS2-1-109 Ansel, Juliet P-26 Bengt A., Andersson DS2-3-08 Antezana, Erick A-08 Benito, Adrián DS1-1-69 Anthony, Mark DS2-4-07 Benjamin, Kirsten DS2-4-04 Aoki, Kazuhiro DS2-1-54 Benkert, Beatrice DS3-3-59 Aperia, Anita DS2-1-65 Benoot, Donald OS-05* Aref, Walid DS2-4-13 Bensch, Robert DS1-4-15 Arendt, Detlev DS1-2-16 Benson, Mikael A-50*, DS2-3-01*, DS2-3-08 Arga, Kazim Yalcin DS2-4-18 Ben-Yehezkel, Tuval P-22 Arkin, Adam DS2-4-24, DS3-1-07 Berg, Johannes DS3-1-10 Armitage, Judith DS3-4-16 Berglund, Martin DS3-3-43 Arne, Gabriella OS-13* Bergmann, Frank DS3-3-40* Asakawa, Takeshi DS2-2-144*, DS2-2-145 Bertrand, Dominique DS2-2-67 Ashall, Louise DS2-1-02 Bertrand, Edouard A-03 Askari, Mohammad Sharif DS1-1-28, DS2-4-06 Berube, Hugo DS1-2-16 Asker, Noomi OS-46 Besozzi, Daniela DS2-1-112 Atari, Mohammed DS1-3-06* Bessières, Philippe A-30

214 ICSB 2008 Bettenbrock, Katja DS1-1-04 Bruggeman, Frank DS1-1-01, DS1-3-01, Bhalla, Upinder S. DS2-2-111 DS2-2-12, OS-02, DS2-1-73, Bhardwaj, Nitin OS-04* DS2-1-81, DS2-1-93 Bhartiya, S DS2-1-106 Brunak, Soeren DS2-1-24 Bhattacharya, Nupur DS2-1-72* Brunner, Michael DS3-4-07 Bhattacharya, Sudin DS2-2-128* Brusch, Lutz DS2-1-98* Bhumiratana, Sakarindr DS1-1-36 Bryzgalova, Nadezda DS2-3-13 Biberdorf, Elina DS2-2-149 Bucher, Elmar DS3-3-32* Birney, Ewan DS1-2-04 Buchler, Nicolas DS3-2-06* Bischofs, Ilka DS2-2-37*, DS2-4-24* Buck, Gregory OS-33 Bittig, Arne T. DS1-2-11, DS2-1-37* Budovsky, Arie DS2-2-34 Bjorkegren, Johan DS2-2-16 Buhmann, Joachim DS3-4-25 Björkegren, Johan DS2-3-02, DS2-3-07, OS-03 Bulik, Sascha DS1-1-35 Blackburn, Hannah DS1-1-08 Burckhardt, Christoph DS2-4-12* Blake, Judy DS3-4-01 Burden, Frank DS2-2-100, DS2-2-127* Blankenship, Derek DS2-1-31 Burke, John DS3-1-09 Blass, Sarah DS2-2-136 Burnett, John DS3-1-07* Blauvelt, Meagan DS1-2-03 Burns, Suzanne DS2-1-11 Blinov, Michael A-38, DS1-2-18 Burroughs, Nigel DS2-2-114* Blits, Marjolein DS1-3-08 Burton, Tarea DS2-1-11 Blodgett, David DS2-3-06 Busch, Hauke DS2-1-72, DS2-1-99, Bloechl, Florian DS2-2-11 DS2-2-13*, DS3-4-07, Blokhin, Alexandr DS2-2-149 DS2-1-40 Blomberg, Anders DS1-1-18, DS2-1-103, Busetto, Alberto Giovanni DS3-4-25* DS2-1-111, DS2-4-45 Busti, Stefano DS2-1-66, DS2-1-71 Blonde, Ward A-08 Byrne, David DS2-2-33* Blust, Ronny OS-05 Byrne, Emma DS3-4-08 Bock, Hans-Georg DS3-4-07 Böcker, Ulrike DS2-4-45* C Bodvard, Kristofer DS2-1-103*, DS3-3-04 Boere, Diana W.M. OS-09* Cai, Yizhi DS1-2-03 Bohl, Sebastian DS2-1-23, DS2-1-99* Calof, Anne L. DS3-1-16, DS2-1-89 Bonin, Michael DS2-1-78 Calvetti, Daniela DS3-3-53 Boone, Charles DS2-4-03, P-04* Calzone, Laurence DS2-1-57* Booth, Ian DS2-2-23, DS2-4-22, Campeanu, Gheorghe DS2-4-33

DS2-4-23, DS2-4-42 Cankorur, Ayca OS-30* Index Author Bordbar, Aarash DS2-2-25 Cardell, Lars Olaf DS2-3-08 Bordel Velasco, Sergio DS1-1-24* Carlberg, Carsten DS2-1-81 Borge, Grethe Iren DS3-3-62 Carneiro, Sónia DS2-2-94* Borger, Simon DS2-2-129 Carroll, Kathleen A-04, DS1-1-65 Bork, Peer DS1-2-16, P-05 Cascante, Marta A-54, DS1-1-69, DS2-3-21* Boronovsky, Stanislav DS2-2-146 Castrillo, Juan I. DS1-1-32* Bortolozzi, Mario DS2-1-97* Castro, Cristiana DS2-1-18, DS2-2-51 Bosch, Daniel DS2-1-109 Catchpole, Gareth DS1-1-50 Bosnacki, Dragan DS2-2-30*, OS-09 Caudron, Maiwen DS2-1-72 Botstein, David P-27* Caudy, Michael DS1-2-04 Bottin, Hélène P-26 Cazzaniga, Paolo DS2-1-112 Boudaoud, Arezki DS1-4-06 Cedersund, Gunnar DS2-1-25*, DS2-2-93 Boudolf, Véronique DS1-4-08 Ceppi, Maurizio DS2-1-31* Bouget, François-Yves DS1-4-27 Chan, Tzu-Min OS-14 Boukh-Viner, Tatiana DS1-1-28, DS2-4-06 Chandra R, Nagasuma DS1-3-25, DS2-2-110 Bourque, Simon DS1-1-28, DS2-4-06 Chang, Andrew DS1-1-28, DS2-4-06 Bouwman, Jildau DS1-1-61, DS1-3-08 Chang, Ivan DS2-2-66* Boyarskikh, Ulyana DS2-1-122 Chang, Yo-Cheng DS2-2-18* Boys, Richard DS2-2-107 Chaouiya, Claudine DS2-1-110, DS2-2-140 Bradshaw, Maria DS2-3-07 Chappell, Michael DS1-3-06, DS3-4-20 Brady, Nathan DS1-3-22 Chassagnole, Christophe DS1-3-05 Brasen, Jens Christian DS1-1-51, DS1-1-57* Chaussabel, Damien DS2-1-31 Brazhe, Alexey DS2-2-86* Chavali, Arvind DS2-4-17* Brazhe, Nadezda DS2-2-86, DS2-3-13* Chazov, Evgeniy DS2-3-13 Brazma, Alvis DS1-2-16 Cheevadhanarak, Supapon DS1-1-07, DS1-1-36 Brenner, Naama DS2-4-39 Chen, Bor-Sen DS2-2-14 Brent, Roger DS2-4-04, DS3-1-05 Chen, Chung-Ming DS2-2-38 Bretschneider, Till DS2-1-09* Chen, Huiyi DS3-1-08 Brezina, V. OS-50 Chen, Luonan DS2-2-84 Brinne, Björn DS2-3-07 Chen, Shu-Hwa DS3-3-20*, DS3-3-22 Brismar, Hjalmar DS2-1-65, DS2-1-67 Chen, Xiao DS1-3-19, DS3-2-14* Brive, Lars DS3-3-56* Chia-Hao, Chin DS3-3-20 Bro, Rasmus DS2-2-72 Chiaradonna, Ferdinando DS2-1-63, DS2-2-96 Broomhead, David A-04, DS1-1-55, DS2-1-02, Chickarmane, Vijay A-37 DS1-1-65 Chin-Wen, Ho DS3-3-20 Brors, Benedikt DS1-3-26 Cho, Chunghee DS1-1-62 Brozek, John A-54 Cho, Hee-Jung DS2-4-40 Brugard, Jan A-54 Cho, Hwan Sung DS2-3-16

ICSB 2008 215 Cho, Kwang-Hyun DS2-1-38, DS2-1-48, de Atauri, Pedro Ramón DS1-1-69* DS2-2-131, DS2-2-22, De Baets, Bernard A-08 DS2-2-62, DS2-2-73, OS-12, de Bono, Bernard DS1-2-04 P-12* De Dieuleveult, Maud P-26 Cho, Sung-Weon OS-45 de Faire, Ulf DS2-3-07 Choi, Jung Kyoon DS2-3-16* de Jong, Hidde DS2-2-07 Choi, Paul DS3-1-08* de la Fuente, Alberto DS2-3-23 Choi, Sangdun DS2-1-102 de Moura, Alessandro DS2-4-42 Choi, Wing DS3-4-05 de Souza Abreu, Raquel DS2-1-11 Chokakthukalam, Achuthanunni DS1-4-07* De Veylder, Lieven DS1-4-08 Choo, Sang-Mok DS2-1-48 De Vink, Erik DS2-2-30 Chou, Cheng-Chung DS2-1-121* de Vos, Willem M. DS2-4-30 Chou, Ching-Shan DS2-1-79, DS2-1-89 de Winde, Han A-06 Chris, Bunce DS2-1-123 de Winde, Johannes H DS1-1-59 Chu, M.-T. DS3-4-11 Degenhardt, Tatjana DS2-1-81 Chuang, Cheng-Long DS2-2-38 DeGioia, Luca DS2-1-71 Chumnanpuen, Pramote DS1-1-10* Deitmer, Joachim DS2-2-63 Ciliberto, Andrea DS2-2-140, OS-02 Del Conte-Zerial, Perla DS2-1-98 Cimpeanu, Carmen DS2-4-33 DeLalla, Emily C. DS1-2-03 Cirulli, Claudia DS2-1-66 Dell, Anne DS2-2-133 Claesson, Marcus DS3-3-47* Dell’Orco, Daniele DS2-1-64 Clare, Amanda DS3-4-08* Delneri, Daniela DS1-1-05 Clarke, Neil DS3-1-13 Dematte’, Lorenzo DS2-2-135 Cline, Melissa DS3-3-58 Denby, Katherine DS1-4-20* Coccetti, Paola DS2-1-66* Deng, Youping DS3-4-13 Coen, Enrico DS1-4-33 Depaulis, Antoine A-02 Collinet, Claudio DS2-1-98 DePauw, Edwin A-02 Colman-Lerner, Alejandro DS2-4-04, DS3-1-02*, Depner, Sofia DS2-1-21* DS3-1-05 Dereli, Elif OS-39* Colombo, Sonia DS2-1-112 D’Eustachio, Peter DS1-2-04 Commichau, Fabian M DS3-3-33 Deutsch, Andreas DS2-1-98 Conolly, Rory B. DS2-2-128 Devine, Kevin A-30

Author Conradie, Riaan A-27, DS1-1-01, DS3-3-07, Dey, Gautam OS-44* Index OS-02* Dhabolkar, Sugat DS3-2-19 Constante, Marco DS3-2-02 Dhondt, Stijn DS1-4-08 Conzelmann, Holger DS3-4-12 di Bernardo, Mario DS3-2-08 Cooper, Krisal L. DS1-2-03 Di Ventura, Barbara DS3-4-07 Corellou, Florence DS1-4-27* Digiuni, Simona DS1-4-03, DS1-4-23 Corominas-Murtra, Bernat DS2-2-64 Dikicioglu, Duygu DS1-1-31, DS1-1-52*, OS-39 CoSBi A-10* Dill, David DS3-4-24 Cosgrove, Benjamin DS2-1-60 Dissanayake, Sarala DS1-2-14 COSMIC Consortium OS-32 Dobrzynski, Maciej DS2-2-12* Costa Martins, Rui DS2-1-18*, DS2-2-51* Dobson, Paul D. OS-38, P-10 Costenoble, Roeland DS1-1-38* Dolowy, Krzysztof OS-37 Couder, Yves DS1-4-06 Dominy, James A-27, DS3-3-44* Cournac, Axel OS-25* Dong, Guangqiang DS3-1-11* Cowan, Ann A-38, DS3-3-05 Dooley, Steven DS2-1-23 Crasta, Oswald DS1-2-03 Douglas, Kell A-51 Croft, David DS1-2-04 Dowson, Christopher DS3-4-20 Csikasz-Nagy, Attila DS2-1-08, DS3-3-01*, OS-02 Doyle III, Francis J DS3-4-06 Curk, Tomaz DS2-1-53 Du Preez, Franco A-27, DS1-1-01, DS3-3-07* Currle, David DS2-1-91 Dubhashi, Devdatt DS2-2-152 Cvijovic, Marija DS2-2-112* Dubrova, Elena DS3-2-11* Cyr, David DS1-1-28, DS2-4-06 Duessmann, Heiko DS3-3-15 Czernik, Dominika DS2-3-24* Dumas Milne Edwards, Jean Baptiste A-02 D Duncan, Hull A-51 Dunn, Warwick A-04, DS1-1-65 da Silva Lopes, Tiago Jose DS1-1-58* Dussmann, Heiko DS2-1-108 Dada, Joseph DS3-3-03, DS3-3-29* Dahl, Peter DS2-1-118 E Dahlgren, Jovanna DS2-3-10 D’Alessandro, Lorenza DS2-1-20*, DS2-1-23 Easton, Douglas DS3-4-01 Dalevi, Daniel DS2-2-139* Ebenhöh, Oliver DS1-4-04, DS1-4-18 Damon, Christelle P-26 Edelstein, Stuart J DS2-1-12 Darabos, Christian DS2-2-46* Ederer, Michael A-28*, DS1-1-04*, DS2-2-03* Daran-Lapujade, Pascale DS1-1-59 Edgren, Henrik DS3-3-32 Dartan, Burcu DS1-4-03, DS1-4-23 Edvardsson, Malin DS1-3-21, OS-17 Darzi, Ara DS2-3-09 Egana, Mikel A-08 Daskalaki, Andriani A-31 Egea, Jose A. DS2-2-71, DS3-3-28 Datsenko, Kirill A. DS2-4-13 Ehrenberg, Måns DS2-4-43 David, McMillen DS3-1-11 Eijken, Marco DS2-1-73 Davies, Mark DS1-3-27* Davis, J DS2-1-55 de Almeida, Camila DS2-4-42

216 ICSB 2008 Eils, Roland A-11, A-42*, DS1-3-22, Fleck, Christian DS1-4-03, DS1-4-11*, DS1-3-26, DS2-1-105, DS1-4-15, DS1-4-23, DS2-1-99, DS2-2-13, DS1-4-25, DS3-4-21 DS3-4-04*, DS3-4-07, Fleming, Ronan M.T. DS2-2-25 DS2-1-40 Florez, Lope A DS3-3-33* Einevoll, Gaute T. DS2-2-117, DS2-2-137 Floris, Matteo DS2-3-23 Ekstrand, Helena OS-17 Flöttmann, Max DS3-3-46* EL Bardicy, Samia DS3-1-15 Flusberg, Deborah DS2-1-51* Elbing, Karin A-23, DS2-1-109 Folkerts, Otto DS1-2-03 Elf, Johan DS1-1-37, DS2-2-79, Font, Julio OS-43 DS3-1-01* Fontes, Magnus DS2-3-22 Elina, Nataliya OS-20 Foreman, Julia DS1-4-05 El-Khayat, Hanaa DS1-1-54 Frahm, Thomas DS2-1-23 Elliot, Elizabeth A-21 Fraifeld, Vadim DS2-2-34 Elowitz, Michael B. P-25* Francis J, Doyle III DS3-4-15 Elvassore, Nicola DS3-4-15 Franco-Cereceda, Anders DS2-3-07 Emig, Dorothea DS3-3-58* Franco-Lara, Ezequiel DS2-2-45 Emili, Andrew DS3-1-08 Franjcic, Zlatko DS2-2-152* Engblom, Stefan DS3-3-42 Franke, Raimo DS2-1-115 Enger, Jonas DS3-1-12 Fraunholz, Martin A-55 English, Brian DS1-1-37* Frey, Simone DS2-1-109*, DS2-1-80*, Eriksson, Emma DS3-1-12* DS3-3-09 Erjavec, Nika DS2-2-112 Friedrichsen, S DS2-1-55 Ermakov, Gennadiy DS1-1-67, DS2-1-96 Frimmersdorf, Eliane DS3-3-59 Ermolov, Andrey DS1-1-27* Fritz, Georg DS3-1-14 Errington, Rachel DS1-3-06 Fröhlich, Martina DS2-1-95* Ertesvåg, Helga DS2-4-34* Fromion, Vincent A-30, DS2-2-105, DS2-2-108 Eskerod Madsen, Bo DS2-3-03 Frost, Rickard DS1-3-21*, OS-17 Eslahpazir, Manely DS2-2-45 Funahashi, Akira DS3-3-12, DS3-3-19*, Espelin, Christopher DS2-1-42 DS3-3-39, DS2-2-93 Ess, Sara DS2-4-13 Fundel, Katrin DS1-3-26 Eufinger, Jan A-11* Furlong, Eileen DS1-2-16, DS2-1-110, P-17* Evans, Clive DS1-2-03 Furman, Itay DS2-2-124 Evans, Neil DS1-3-06, DS3-4-20 Ewald, Jennifer P-13 G Index Author F Gaglio, Daniela DS2-2-96 Gaidatzis, Dimos DS2-1-76 Faergestad, Ellen Mosleth DS2-2-72 Galinina, Nina DS1-1-14 Faizi, Amir DS1-4-32 Gao, Yuan OS-33 Fajardo Sanchez, Emmanuel DS3-2-02 Garapati, Phani DS1-2-04 Falciani, Francesco A-54 Garcia de Carvalho, Andreia DS3-2-02 Falck, Eva OS-22* Garcia-Salcedo, Raul DS2-1-109 Fanelli, Francesca DS2-1-64 Gardner, Timothy DS2-2-33 Fange, David DS2-4-43* Garla, Vijay DS1-1-56* Fantinato, Sonia DS2-1-71 Garriga Canut, Mireia DS3-2-02 Faratian, Dana DS1-3-24 Gaskell, Simon J. DS1-1-65, A-04 Farina, Lorenzo DS1-1-41 Gaudet, Suzanne DS3-4-03 Farina, Marcello DS2-1-71 Gauges, Ralph DS3-3-03 Faure, Adrien DS2-1-110, DS2-2-140* Gavin, Anne-Claude DS2-1-56 Fehrmann, Steffen P-26 Gawish, Fathyia DS1-1-54*, DS3-1-15* Feiler, Heidi DS2-1-70 Gayen, Kalyan DS3-4-06 Fein, Marc DS2-2-48* Gebhardt, Rolf DS2-1-23 Feizi, Amir DS1-4-31 Geier, Florian DS1-4-03*, DS1-4-23*, Fell, David DS1-3-05, DS1-4-07, DS3-4-21* DS2-2-122*, DS2-4-07, Gelay, Amélie DS2-1-57 DS1-1-25 Gendelman, Rina DS2-1-70* Fendt, Sarah-Maria DS1-1-03*, P-13 Gennemark, Peter DS1-2-10* Ferm, Lars DS3-3-42 Gerber, Susanne DS2-2-129* Fernandez, Elia OS-43 Gerland, Ulrich DS3-1-14 Fernandez, Eric DS1-3-05 Gerlee, Philip DS1-1-33* Ferreira, Eugénio DS2-2-94 Gerstein, Mark OS-04 Fey, Vidal DS3-3-32 Gevorgyan, Albert DS2-2-122, DS2-4-07* Filipenko, Maxim DS2-1-122 Geyer, Tihamer DS2-2-136* Filippis, Ioannis DS2-2-39 Ghanegaonkar, Shashank DS2-4-09 Finkenstadt, B DS2-1-55 Ghosh, Amit OS-42* Fiorani, Fabio A-12* Ghosh, Samik DS2-1-77, OS-11*, DS2-1-34 Fioretos, Thoas DS2-3-22 Ghraham, Alison DS2-2-89 Fischer, Bernd DS3-4-25 Giacobini, Mario DS2-2-46 Fischer, Hans-Peter A-30 Gianchandani, Erwin DS1-1-15* Fisher, Paul DS3-3-25 Giese, N. DS2-1-40 Fitzgerald, J. B. P-19 Giese, Nathalia DS2-2-13 Flamm, Christoph DS1-1-29 Gigante, Bruna DS2-3-07 Giladi, Nir DS2-3-05

ICSB 2008 217 Gilles, Ernst Dieter DS1-1-04, DS2-2-03, Guthke, Reinhard DS2-2-68 DS3-3-16, DS3-3-17, DS3-3-18, DS3-4-12, H DS2-1-115 Gillespie, Marc DS1-2-04 Haanstra, Jurgen DS1-3-08 Ginkel, Martin DS3-3-17, DS3-3-18 Haapa-Paananen, Saija DS3-3-32 Giuliani, Alessandro DS2-1-102, DS2-2-87 Habib, Tanwir DS3-4-13* Gjuvsland, Arne DS2-2-120 Haddad, Isam DS3-3-59* Gkargkas, Konstantinos DS2-4-15* Hafner, Marc DS2-2-104* Glaser, Walter DS2-4-14 Hägg, Sara DS2-3-07* Godoy, Patricio DS2-1-23 Hagmar, Jonas DS2-2-08, DS3-3-04 Goelzer, Anne DS2-2-105, DS2-2-108* Hahn-Zoric, Mirjana DS2-3-08 Goemann, Björn DS2-2-83 Halley, Julianne DS2-2-127, DS2-2-100* Goerlitz, Linus DS1-3-04 Halliday, Karen DS1-4-05* Goertsches, Robert DS2-2-68 Halperin, Eran OS-06 Gokoffski, Kimberly DS2-1-89, DS3-1-16* Halstead, Matt DS1-2-14 Goksör, Mattias DS3-1-12 Hamant, Olivier DS1-4-06 Golda, Rostyslav DS3-3-31 Hamsten, Anders DS2-3-07 Goldberg, Alexander DS1-1-28, DS2-4-06 Han, Beomsoo DS3-3-26 Goldsipe, Arthur DS2-1-42* Hanafi, Mohamed DS3-3-62 Golebiewski, Martin DS1-2-05* Hans, Söderlund A-49 Golks, Alexander DS3-4-04 Hansen, Kim DS1-1-42 Goltsov, Alexey DS1-3-24 Hansen, Michael Adsetts Edbe DS1-1-10 Gómez, Federico DS2-3-21 Hardiman, Timo DS1-1-21* Gonçalves, Joana P. DS3-3-13 Härdin, Hanna DS2-2-148* González, Adonaí DS2-2-76 Hardy, Adam DS1-3-05 González, José DS2-2-76 Harms, B.D. P-19 Gonzalez-Lergier, Joanna DS2-1-62* Harper, Claire DS2-1-02, DS2-1-55* Gopinath, Gopal DS1-2-04 Harrison, David DS1-3-24 Gorban, Alexander DS2-2-106 Harrison, Richard DS1-1-05 Gordon, Andrew DS2-4-04 Harwood, Colin A-30 Gordon, Sean A-37 Hasan, Md. Nabiul DS2-2-151*

Author Gormley, Padhraig DS2-2-132* Hasdemir, Dicle DS2-1-47* Index Goryanin, Igor DS1-1-17*, DS1-1-23, Hashemi, Sayed Hossein DS1-1-18 DS1-3-24, DS1-4-05, Hassani, Sahar DS3-3-62* DS2-4-11 Hatakeyama, Mariko DS2-1-116, DS2-1-27 Gotoh, Noriko DS2-1-119, DS2-1-27, Hatanaka, Yousuke DS2-1-50 DS2-1-50 Hattori, Seisuke DS2-1-117, DS2-1-119 Gotti, Laura DS2-1-66 Haudry, Yannick DS1-2-16* Govaerts, Willy DS1-4-08, DS3-3-36* Haus, Sylvia DS2-2-77* Gowrishankar, Swetha OS-44 Hausser, Jean DS2-1-76* Grandori, Rita DS2-1-66 Havecker, Ericka OS-20 Grãos, Mário DS3-3-13 Havrylov, Serhiy DS2-1-94* Gray, Alex DS2-2-118 Hayama, Akemi DS3-3-45 Gray, Joe DS2-1-70 Hayashi, Kentaro DS2-1-32* Grebenisan, Irina DS2-4-33* Hayes, Andrew DS1-1-32 Greber, Urs DS2-4-12 Hayes, Andy DS1-1-31, DS1-1-52 Grebogi, Celso DS1-1-13, DS2-2-23, Hayes, Wayne DS2-1-91 DS2-2-81, DS2-4-23 Haynes, Ken DS2-1-19 Greco, Claudio DS2-1-71 He, Feng DS2-1-52* Green, Jeff DS2-2-89 Heap, John T. DS2-1-107 Greenberg, David A-02 Hecker, Michael A-30, DS2-2-68* Greese, Bettina DS1-4-03, DS1-4-15*, Heidari, Mostafa DS1-4-10* DS1-4-23 Heidelberger, Maike A-48* Gregg, Christopher DS1-1-28, DS2-4-06 Heijne, Wilbert DS2-4-38* Gretz, Norbert DS2-1-23 Heijnen, Sef A-06 Gribskov, Michael R. DS2-4-13 Heimstaedt, Susi OS-20 Grimbs, Sergio DS1-4-14 Heinemann, Matthias DS1-1-46, DS2-2-91, Grotli, Morten DS2-1-03 DS3-1-19 GSSB, all groups A-20 Heino, Jenni DS3-3-53* Guda, Sanjay Kumar DS2-2-123* Heinzle, Elmar DS1-3-09*, DS2-4-21 Güell, Marc DS1-1-47, DS2-1-56* Heisler, Marcus A-37, DS1-4-06 Guk, Sang Mi OS-45 Heisner, Ute A-33* Gull, Keith DS2-4-44 Hellander, Andreas DS3-3-42* Gunka, Katrin DS3-3-33 Hellgren, Gunnel DS2-3-10* Gunnarsson, Lina OS-40 Hellgren, Lars I. DS1-1-20 Günther, Ulrich DS2-3-21 Hellgren, Mikko DS1-1-22* Guo, Guoji DS3-1-13 Hellwig, Christian DS2-1-108 Gupta, Gagan OS-44 Helms, Volkhard DS2-2-136 Gupta, Raj DS3-1-17 Helmy, Mohamed DS2-1-32, DS2-1-90* Gursky, Vitaly DS2-2-143*, DS3-3-51*, Hempel, Dietmar Christian DS2-2-45 OS-16* Henderson, Daniel DS2-2-107 Guseva, Ksenia DS1-1-13* Hengl, Stefan DS2-1-101, DS2-1-17, Gustafsson, Mika DS2-2-16* DS2-1-26, DS2-1-33, Gustafsson, Thomas OS-49 DS2-2-02

218 ICSB 2008 Hengstler, Jan G. DS2-1-23, DS2-1-20 Hughes, Timothy DS2-4-03 Henney, Adriano DS1-3-02* Huisinga, Wilhelm DS1-3-07, DS1-3-23 Henrich, Thorsten DS1-2-16, DS2-1-28 Hulskamp, Martin DS1-4-03, DS1-4-23, Herrmann, Frank DS3-2-02 DS1-4-15 Herwig, Ralf A-31 Hundeshagen, Phillip DS1-3-22* Herzel, Hanspeter DS2-1-23, DS2-1-36 Hunte, Kerrilyn DS2-3-09 Heslop-Harrison, Pat DS2-1-10, OS-12 Hunter, Peter DS1-2-09, P-18* Hess, Wolfgang R. DS2-4-28 Hurst, Laurence DS1-1-39 Heuer, Andreas DS1-2-12 Huss, Mikael DS3-1-13* Heux, Stephanie P-13 Hussong, Andrea DS1-4-11 Higham, Desmond, J. DS2-2-41 Huuskonen, Anne DS1-1-68 Hilbers, Peter DS1-1-06, DS1-1-48, Hwa, Terence DS2-4-19, DS2-4-36, DS2-1-15, DS2-1-41, DS3-2-05* DS2-2-56, OS-09 Hill, Anthony DS3-2-07 I Hill, David DS3-4-01 Hiller, Karsten DS3-3-59 Ibig, Ariane DS2-2-09* Hiort, Catharina A-18 Iborra, Francisco DS3-1-17 Hirmajer, Tomas DS3-3-48 Ichinose, Junya DS2-1-05 Hirsch, Alison DS3-4-17 Ikeda, Kazuhiro DS2-1-27 Hitchen, Paul DS2-2-133 Iljin, Kristiina DS3-3-32 Ho, Chun-Ming DS2-3-04 Imoto, Seiya DS2-1-50 Hochberg, Ze‘ev DS2-3-10 Imreh, Gabriela A-05* Hodgman, Charlie A-29 Inoue, Jun-ichiro DS2-1-117, DS2-1-119 Hoefer, Thomas DS2-1-20, DS2-1-86, P-15 Inoue, Satoshi DS2-1-116, DS2-1-27 Höfer, Thomas DS2-2-121, DS2-2-32* Ioannis, Lestas DS3-1-06 Hoffmann, Christian DS3-4-07 Isaeva, Julia DS2-2-72* Hoffmann, Sabrina DS2-2-134, DS2-2-138*, Isalan, Mark DS3-2-02* DS1-1-35 Ishido, Yusuke DS2-1-16 Hofmann, Ute DS1-3-03 Ishihara, Shuji DS2-1-92 Hofmeyr, Jan-Hendrik DS3-3-23, DS3-3-35, Ishii, Shin DS2-1-54 DS3-3-44, DS2-2-52* Ito, Caryn DS2-1-07 Hofmeyr, Jannie A-27 Ito, Masahiro DS2-1-16* Hohmann, Stefan A-17*, A-18*, A-19*, A-20*, Ittrich, Carina DS1-3-26 A-23*, A-24*, DS2-1-03, Ivanic, Joseph DS2-2-70

DS2-1-109, DS2-1-111, Ivanova, Lyudmila DS2-2-149 Index Author DS2-1-118, DS2-1-66, Ivanova, Natalia DS2-2-139 DS2-1-80 Ivert, Torbjörn DS2-3-07 Holcombe, Mike DS2-2-89, DS3-3-25 Iyer, Kavita DS3-2-07 Holl, Mark DS2-4-04 Izaguirre, Jesús DS3-3-37 Hollmann, Susanne A-33 Holm, Kora A-27 J Holman, Tara A-29* Holmes, Elaine DS2-3-09 Jabbari, Sara DS2-1-107* Holmes, Raquell A-38 Jablonsky, Jiri DS1-4-16* Holstege, Frank P-03* Jackers, Pascale A-02 Holzhuetter, Hermann-Georg DS2-2-134, DS2-2-138, Jackson, Kim DS1-1-08 DS1-1-35 Jacobsen, Elling DS2-1-126 Honda, Hiroyuki DS2-3-19 Jacobsen, Elling W. DS1-1-22, DS3-4-02 Hong, Christian I DS2-1-08 Jacobsen, Marc OS-18 Hong, Seong-Eui DS1-1-62, DS2-1-35* Jacobson, Ingemar OS-10 Hood, Lee P-29* Jacobson, Therese DS2-1-118 Hoog, Jan-Olov DS1-1-22 Jacquet, Claire DS1-4-27 Hoops, Stefan DS2-2-28, DS3-3-03 Jahn, Dieter DS3-3-59 Hoppe, Andreas DS1-1-35, DS2-2-134* Jain, Pragati DS2-2-111* Hornquist, Michael DS2-2-16 Jakobsen, Øyvind M. DS2-4-34 Horton, Caroline DS2-1-02 James, Thorne DS2-1-123 Hou, Jin DS1-1-60* Jameson, Daniel A-04, DS1-2-15* Houtsmuller, Adriaan P-15 Jansen, Stefan DS2-4-28 Hovig, Eivind DS2-1-58, DS2-2-150 Jarmer, Hanne Ø. A-30 Hoyle, David DS1-1-32 Jassal, Bijay DS1-2-04 Hrabé de Angelis, Martin DS2-1-88 Jauhiainen, Alexandra OS-40* Hryshchenko, Nataliya DS2-3-17* Jeffries, Thomas W. OS-24 Hsin-Hung, Wu DS3-3-20 Jeneson, Jeroen DS1-1-06, DS1-1-48 Hsu, Chao-Ping DS3-4-11* Jeon, Noo Li DS2-1-79 Hsu, Jia-Wei DS1-3-18 Jin, Sora DS1-1-62 Huang, Hsuan-Cheng DS1-3-18 Jin, Yaochu DS2-2-147 Huang, KC DS2-4-16* Jirstrand, Mats DS2-1-103, DS2-2-08, Huang, Le DS1-3-19 DS2-2-63, DS3-3-08*, Hubbard, Simon J DS3-3-55 DS3-3-43, OS-10 Huber, Heinrich DS2-1-108, DS3-3-15 Johansson, Cecilia DS2-1-25 Hucka, Michael A-53*, DS3-3-50 John, Mathias DS1-2-12* Hug, Josh DS2-4-24 Jones, Nick A-35*, DS2-2-119* Hugenholtz, Jeroen A-06, DS1-1-16 Jönsson, Henrik A-37*, DS1-4-30, DS1-4-06 HugenHoltz, Phil DS2-2-139 Joo, Kyung-Whan OS-45

ICSB 2008 219 Joseph, Simpson DS2-4-19 Kemp, Melissa DS3-4-17 Joshi, Himanshu N DS2-2-150* Kemper, Brian DS3-3-52 Jou, Yuh-Shan DS2-3-04 Kerkhoven, Eduard DS1-3-08 Jouhten, Paula DS1-1-66, DS1-1-68* Kern, Sabine OS-18 Jouraku, Akiya DS3-3-19, DS3-3-39 Kerr, Lorraine A-21* Jozefczuk, Szymon DS1-1-50* Khalil, Iya DS2-1-70 Juan, Hsueh-Fen DS1-3-18* Khan, Farid A-04, DS1-1-65 Juarez, Miguel DS2-2-114 Kheibarshekan ASL, leila DS1-4-08*, DS3-3-36 Juhl Jensen, Lars DS2-1-24, P-05 Kholodenko, Boris DS2-1-120 Julenius, Karin OS-07* Khrytankova, I. OS-50 Jun’ichi, Tsujii A-51 Kiel, Christina DS2-1-46* Jung, Alexander A-30 Kielland-Brandt, Morten DS2-1-104 Jung, Kirsten DS3-1-14 Kihara, Daisuke DS2-4-13 Jung, Sung Hoon DS2-2-22, OS-12 Kikuchi, Norihiro DS3-3-19, DS3-3-45, Jungreuthmayer, Christian DS2-4-31 DS3-3-52 Junion, Guillaume DS2-1-110 Kille, Peter DS2-2-118 Juvan, Peter DS1-1-44 Kilpinen, Sami DS3-3-32 Jyrki, Lötjönen A-49 Kim, Chul Whan OS-45 Kim, Do Han DS1-1-62, DS2-1-35 K Kim, Dongsan DS2-2-62* Kim, Hyun Sook OS-45 K.R., Vivek Sagar DS2-2-123 Kim, Hyun Young DS1-1-28, DS2-4-06 Köhn, Dagmar DS1-2-12 Kim, Hyunchul OS-21* Kaandorp, Jaap DS3-3-51 Kim, Hyungtae DS2-4-41 Kaderali, Lars DS2-2-60* Kim, Jeong-Gu DS2-4-40 Kaferle, Petra DS2-1-53 Kim, Jeong-Rae DS2-2-73*, OS-12 Kahm, Matthias DS2-2-129 Kim, Jongrae DS2-1-10* Kainth, Pinay DS2-4-03 Kim, Junil OS-12* Kaisa, Poutanen A-49 Kim, Kyung H. DS2-2-113* Kaiser, Dale DS3-3-37 Kim, Sangtae DS2-4-13 Kaizu, Kazunari DS1-3-12, DS3-4-22*, OS-11 Kim, Tae-Geon OS-12 Kalaidzidis, Yannis DS2-1-98 Kim, Tae-Hwan A-26*, DS2-2-22*

Author Kalamatianos, Dimitris DS3-3-15 Kim, Young-Joon DS2-3-16 Index Kalapanulak, Saowalak DS2-4-11* King, John R. DS2-1-107 Kalko, Susana A-54 King, Ross D. DS3-4-08 Kall, Mikael DS1-1-18, DS2-1-103, Kinross, James DS2-3-09* DS3-3-04 Kircher, Stefan DS1-4-11, DS1-4-25 Kallioniemi, Olli DS3-3-32 Kirdar, Betul DS1-1-31, DS1-1-52, Kalnenieks, Uldis DS1-1-14* DS2-1-47, OS-30, OS-39 Kamali-Zare, Padideh DS2-1-65*, DS2-1-67 Kirkwood, Tom A-46, DS2-2-107 Kamburov, Atanas A-31 Kirouac, Dan DS2-1-07* Kaminski, Norbert E. DS2-2-128 Kiselev, Ilya DS3-3-60 Kamp, Christel DS2-2-42* Kishony, Roy DS2-4-05, P-09* Kanai, Akio OS-21 Kitajima, S DS2-1-34 Kanapin, Alexander DS1-2-04 Kitano, Hiroaki A-44*, DS1-3-12, DS2-1-111, Kaneko, Kunihiko DS2-1-05 DS2-1-116, DS2-1-27, Kania, Renate DS1-2-05 DS2-1-34, DS2-1-77, Kanno, J DS2-1-34 DS2-2-93, DS3-3-12, Kaplan, Shai P-22 DS3-3-19, DS3-3-39, Kapuy, Orsolya P-14 DS3-3-45, DS3-3-52, Karhumaa, Kaisa DS2-1-104* DS3-4-22, OS-11, P-28* Karschau, Jens DS2-4-42* Kitanovic, Ana DS2-4-14 Kasemo, Bengt DS1-3-21, OS-17 Klamt, Steffen DS2-1-06, DS2-1-115, Kashiwagi, Akiko DS2-1-114 DS2-2-69, DS3-3-54 Kaski, Sami DS2-2-85 Klann, Michael DS2-2-31* Kaski, Samuel DS1-1-27 Kleijn, Roelco DS1-1-66 Katajamaa, Mikko DS1-1-02, DS2-3-15 Klein, Dagmara DS2-1-118 Kay, Emily DS2-2-133 Klein, Franziska DS2-2-37 Kayembe, Mulanda DS1-1-28, DS2-4-06 Klein, Johannes DS2-4-32* Kazak, Hande DS2-4-18 Klinga Levan, Karin OS-22 Kaznessis, Yiannis DS3-2-07* Klingmüller, Ursula A-11, DS2-1-01*, DS2-1-17, Kazuta, Yasuaki DS3-4-10 DS2-1-23, DS2-1-33, Ke, Ruian DS2-1-19* DS2-1-101, DS2-1-20, Keating, Sarah DS3-3-50* DS2-1-26, DS2-1-86, Kedmi, Merav DS2-3-05 DS2-1-99 Keienburg, Jens DS3-4-07* Klipp, Edda A-30, DS2-1-03, DS2-1-118, Kekow, Joern DS2-2-68 DS2-1-95, DS2-2-01*, Kel, Alexander DS1-2-19, DS2-1-74* DS2-2-112, DS2-2-129, Kell, Douglas A-04, DS1-1-30, DS1-1-63, DS2-2-130, DS2-2-19, DS1-2-15, DS2-1-02, DS2-2-95, DS2-2-96, DS3-3-49, DS3-3-55, DS2-2-97, DS3-3-21, DS3-3-61, P-10*, DS1-1-55, DS3-3-24, DS3-3-46 DS1-1-65, DS2-4-14, OS-38 Klumpp, Stefan DS2-4-36 Kelly, Steven DS2-4-44* Knabe, Johannes F. DS2-2-103 Kel-Margoulis, Olga A-02 Knapen, Dries OS-05

220 ICSB 2008 Knudsen, Michael DS2-2-04 L Knuuttila, Juha E.A. DS2-2-85 Ko, Daijin DS2-1-11 Lagzdina, Laura DS1-1-64 Kobraee, Sohil DS1-4-26 Lahti, Leo DS2-2-85* Koch, Ina DS2-2-125* Lai, Jason DS3-3-11 Koczan, Dirk DS2-2-68 Laing, Emma DS3-3-55* Koehler, Gottfried DS2-1-87 Lamanna, Charles DS3-3-37 Koehn, Dagmar DS1-2-01 Lambin, Philippe DS2-1-41 Koeppl, Heinz DS2-2-104 Lapukhov, Sergey DS3-3-60 Kohlbacher, Oliver DS2-1-14 Lämmle, Bärbel DS1-3-26 Kohler, Achim DS2-4-45, DS3-3-62 Lander, Arthur DS2-1-89*, DS2-1-91, Kohlwein, Sepp D. DS2-4-31 DS3-1-16 Kohno, Takashi DS2-1-50 Landthaler, Markus DS2-1-76 Kohonen, Pekka DS3-3-32 Lang, Patrick DS2-2-63 Koike, Nobuya DS3-4-18 Langdon, Simon DS1-3-24 Koivumäki, Jussi DS2-1-61* Langston, Michael A. DS2-3-08 Koizumi, Moriyoshi A-52* Lanthaler, Karin OS-38* Kolch, Walter DS2-1-17 Laoteng, Kobkul DS1-1-07, DS1-1-36 Kolodkin, Alexey DS2-1-93* Lapin, Alexei DS2-2-31 Kolpakov, Fedor DS1-2-19, DS2-1-125*, Lappe, Michael DS2-2-39 DS2-1-96, DS2-2-149, Lars, Förlin OS-46 DS3-3-60* Larsson, D.G. Joakim OS-40, OS-46 Komamura-Kohno, Yuki DS3-4-18 Lauck, Florian DS2-2-136 Kondrakhin, Yuriy DS2-1-122 Lauffenburger, Douglas DS2-1-06, DS2-1-42, Konrad, Peter DS2-3-07 DS2-1-60, DS3-4-03, P-07*, Korhonen, Topi DS2-1-61 DS3-4-19 Korn, W. Michael DS2-1-70, DS3-4-19 Lavrik, Inna DS2-1-105*, DS3-4-04 Kostic, Alex DS2-4-03 Lawson, James DS1-2-09 Kotte, Oliver DS2-2-91* Lazar, Dusan DS1-4-16 Kovacs, Karoly DS1-1-39* Le Novere, Nicolas A-36*, DS1-2-01*, DS1-2-13, Kowalewski, Jacob M. DS2-1-65, DS2-1-67* DS2-1-12 Kozlov, Konstantin DS3-3-51 Leander, Karin DS2-3-07 Kozlowski, Cleopatra P-16 Lebedeva, Galina A-21, DS1-3-24* Kozuka-Hata, Hiroko DS2-1-116, DS2-1-117*, Lebiedz, Dirk DS2-2-47*, DS2-2-48 DS2-1-119 Lee, Baek-Seok OS-21

Kramer, Achim DS2-1-36 Lee, Byoung-Moo DS2-4-40*, DS2-4-41 Index Author Krammer, Peter H. DS3-4-04, DS2-1-105 Lee, David DS2-4-24 Krantz, Marcus DS2-1-111, DS2-2-93* Lee, Hyeon-Woo DS2-2-131 Krause, Falko DS3-3-24 Lee, Hyunju DS3-3-26* Kravchenko, Sergey DS2-3-20 Lee, Jeong-Bong OS-27 Krebs, Olga DS1-2-05 Lee, Joon-Sang OS-45 Kreft, Jan-Ulrich DS2-4-39 Lee, Ju Hyun OS-45* Kremling, Andreas A-47*, DS1-1-53*, DS2-2-61, Lee, Keun Woo DS2-1-48 DS2-2-71, DS2-4-42 Lee, Pohan DS2-4-36* Kreutz, Clemens DS2-2-02 Lefranc, Marc DS1-4-27, DS2-2-153* Krippendorff, Ben-Fillippo DS1-3-23* Legewie, Stefan DS2-1-23 Kristiansson, Erik DS2-2-139, OS-13, OS-40, Legname, Giuseppe DS2-3-11 OS-46* Lehmann, Irina A-11 Krogsgaard, Steen DS1-1-42 Lehmann, Wolf D. DS2-1-17 Krupinski, Pawel A-37, DS1-4-06* Lehner, Ben OS-48 Kschischo, Maik DS2-2-129 Lehnert, Dirk DS2-2-37 Kuchinsky, Allan A-07 Lehrach, Hans A-31 Kuchler, Karl DS1-1-63, DS2-4-14 Lelli, Andrea DS2-1-97 Kudla, A. J. P-19 Lemaire, François DS2-2-153 Kudla, Grzegorz DS3-2-04* Lemerle, Caroline DS3-2-02 Kuepfer, Lars DS1-1-46, DS1-3-04* Lemuth, Karin DS1-1-21, DS2-4-09* Kühn, Clemens DS2-1-118 Lengauer, Thomas DS3-3-58 Kuiper, Martin A-08* Leonova, Tatyana DS2-2-149 Kullingsjo, Johan DS1-4-21* Leroy, Prune DS1-1-37 Kulms, Dagmar DS2-2-90 Lesk, Victor DS2-2-133 Kumagai, Yutaro DS2-1-85* Letellier, Thierry DS2-2-66 Kumamoto, Hiromi DS2-2-101 Leupold, Stefan DS2-4-32 Kumar, A. P-19 Levitner, T. OS-50 Kummer, Ursula DS2-2-28, DS3-3-03 Lewis, Peter J. A-30 Kunze, Angelika DS1-3-21, OS-17 Li, Chen DS1-2-13* Kurat, Christoph F. DS2-4-31 Li, Cheng-Wei DS2-2-14* Kuroki, Yoko DS2-1-116, DS2-1-27 Li, Gene-Wei DS3-1-01 Kuwahara, Hiroyuki A-45, DS3-3-02 Li, Kang DS2-2-132 Kuzu, Guray DS1-1-31* Li, Sabrina DS3-2-05 Kvarnström, Mats DS2-2-08, DS3-3-04*, Li, Ye DS3-3-05 DS2-1-103 Liakata, Maria DS3-4-08 Kwon, Juntae DS1-1-62 Liberati, Diego DS2-1-63, DS2-1-71 Kwon, Ki-Sun DS2-1-48 Lichtenberg-Fraté, Hella DS2-2-129 Kwon, Yung-Keun DS2-2-62 Lichter, Peter DS2-1-72 Kyryakov, Pavlo DS1-1-28, DS2-4-06 Lickert, Heiko DS2-2-109

ICSB 2008 221 Liebal, Ulf W. DS2-2-40 Maksimov, Georgy DS2-3-13 Liebermeister, Wolfram DS2-2-01, DS3-3-24* Malboobi, Mohammad Ali DS1-4-31*, DS1-4-32*, Likhovidova, Elena DS1-2-19, DS2-1-125 DS1-4-32 Liland, Kristian H. DS2-2-72 Maleki, Shohreh DS2-3-02, DS2-3-07 Lilienbaum, Alain DS2-2-106 Maleki-Dizaji, Saeedeh DS2-2-89*, DS3-3-25* Lilljebjörn, Henrik DS2-3-22 Malys, Naglis A-04, DS1-1-30, DS1-1-65 Lin, Chung-Yen DS3-3-20, DS3-3-22* Mammano, Fabio DS2-1-97 Lind, Kristina DS2-1-49 Mandal, Mahuya DS3-1-11 Lindahl, Paul DS2-2-54* Mane, S.P. DS1-2-03 Lindfors, Erno OS-06 Manshaei, Roozbeh DS1-4-31, DS1-4-32 Linding, Rune DS2-1-24*, P-05* Manu DS2-2-143 Lindsey, Keith DS1-4-13 Mao, Xuerong DS2-1-10 Linzhiz, Gregory P-22 Mariko, Hatakeyama DS2-2-144 Lippert, Joerg DS1-3-04 Markel, Arkadiy DS2-2-149 Liska, Jan DS2-3-07 Markham, Magdalena DS3-4-08 Liu, Aiwen DS2-4-24 Markowitz, Victor DS2-2-139 Liu, Jian-Qin DS2-2-141* Markström, Martin A-17, A-18 Liu, Junli Liu DS1-4-13* Marr, Carsten DS2-2-109 Liu, Shumo DS3-2-05 Marsalek, J. OS-50 Liu, Xin DS2-4-36 Martegani, Enzo DS2-1-112* Livshits, Ludmila DS2-3-17 Martens, Harald DS2-2-120*, DS2-2-67*, Livshyts, Ganna DS2-3-20* DS2-2-72, DS2-4-45, Lloyd, Catherine DS1-2-09*, DS1-2-14* DS3-3-62 Lo, Wing-Cheong DS2-1-89 Martens, Magni DS2-2-67 Lockowandt, Ulf DS2-3-07 Martin, Ginkel DS3-3-16 Loew, Leslie A-38, DS3-3-05 Martin, P. DS1-4-21 Logg, Katarina DS1-1-18*, DS2-1-103, Martins, Ana M. DS2-2-98 DS3-3-04 Maruyama, Kozue DS2-1-16 Lohrasebi, Tahmineh DS1-4-31, DS1-4-32 Maschke-Dutz, Elisabeth A-31 Lojek, Agnieszka OS-37* Mashkovtseva, Elena DS2-2-146* Lok, Larry DS2-4-04 Masuoka, Yukiko DS3-3-12 Lopes, Vitor DS2-1-18, DS2-2-51 Mateescu, Eddie DS2-4-19

Author Lopez-Aviles, Sandra P-14 Matsuda, Tetsuya DS3-3-30 Index Lötstedt, Per DS3-3-42 Matsuoka, Yukiko DS2-1-34, DS2-1-77, Lourenço, Anália DS2-2-94 DS3-3-19, DS3-3-45*, Lu, Hang DS3-4-17 DS3-3-52, OS-11 Lu, James DS2-1-87*, DS2-2-26 Matsuura, Tomoaki DS3-4-10* Ludwig, Jost DS2-2-129 Matsuzaki, Jun DS1-4-09* Luijsterburg, Martijn S. DS2-2-121, P-15 Matsuzaki, Yuri A-52 Lundh, Torbjörn DS1-1-33 Matthews, Lisa DS1-2-04 Lundström, Jesper DS2-3-07, OS-03* Mattiazzi, Mojca DS2-1-53 Luneva, Oksana DS2-3-13 Mauch, Klaus DS1-3-03 Luni, Camilla DS3-4-15* Mauri, Giancarlo DS2-1-112 Luo, Kunxin DS2-1-23 May, Patrick DS1-4-18* Luo, Wenlong DS3-1-13 Mayor, Satyajit OS-44 Lutter, Dominik DS2-2-109* Mazein, Alexander DS1-1-17 Lydall, David A-46 Mazur, Johanna DS2-2-60 Lydon, Susannah A-29 Mbodj, Abibatou DS2-1-110* Lyons, Blair M. DS1-2-03 McCarthy, John E.G. A-04, DS1-1-65 Lyu, Jaemyun DS2-3-16 McMillen, David DS3-2-13 McSharry, Patrick DS3-4-16 M Medina, Miguel Angel DS1-1-12 Medina-Gomez, Gema DS1-1-02, DS2-3-15 Ma, Hongwu DS1-1-23*, DS2-4-11 Medrala, Darmara DS2-1-03 Ma, Maggie DS2-3-06 Meechai, Asawin DS1-1-36 Maaheimo, Hannu DS1-1-66, DS1-1-68 Megerle, Judith DS3-1-14* Mach, Valerie DS1-4-03, DS1-4-23 Mehrpanah, Hamid DS1-4-26 Machkalyan, Gayane DS1-1-28, DS2-4-06 Melke, Pontus A-37 Machlica, L. OS-50 Melzer, Guido DS2-2-45, DS2-2-45* Machne, Rainer DS2-1-87, DS2-2-26 Mendes, Pedro A-04, DS1-1-65, DS2-2-28, Maciejewski, Paul DS1-1-09* DS3-3-03*, DS3-3-29, MacLean, Dan OS-20* DS3-3-61 Maclear, Athlee S. DS1-1-11* Mendes, Rui DS2-2-94 Madsen, Curtis A-45, DS3-3-02 Mensonides, Femke DS1-1-63*, DS2-4-14 Mærk, Mali DS2-4-34 Mentzen, Wieslawa DS2-3-23* Maeshiro, Tetsuya DS3-3-38* Merajver, Sofia D DS2-1-22 Maier, Dieter A-54 Merja, Penttilä A-49 Maier, Klaus DS1-3-03* Mertens, Daniel DS2-1-72 Maier, Tobias DS1-1-47 Messiha, Hanan DS1-1-30, DS1-1-65, Maini, Philip DS3-4-16 DS2-4-14 Maiwald, Thomas DS2-1-17, DS2-1-20, Mewes, Hans-Werner A-11 DS2-1-23, DS2-1-26, Meyer, Christoph DS2-1-23 DS2-2-02, DS3-3-06* Meyerowitz, Elliot A-37, DS1-4-06 Makanae, Koji DS1-3-12 Mi, Huaiyu DS3-3-12* Makela, Rami DS3-3-32 Michel, Jean-Baptiste DS2-4-05*

222 ICSB 2008 Michels, Paul DS1-3-08 Myasnikova, Ekaterina OS-16 Middleton, Alistair A-29 Myers, Chris J. A-45*, DS3-3-02* Mikhailov, Alexander S. DS2-2-36 Milanesi, Luciano DS2-2-96 N

Milenkovic, Tijana DS2-2-116, DS2-2-39, N, Hemanth Kumar DS2-2-123 DS3-3-11 Nacher, Jose C DS1-3-13 Milijevic, Svetlana DS1-1-28, DS2-4-06 Nagarajan, M. P-26 Millar, Andrew J. DS1-4-12, DS1-4-05 Nagasaki, Masao DS2-1-119, DS2-1-50 Millard, Bjorn DS2-1-51 Nagashima, Takeshi DS2-1-116, DS2-1-27*, Millat, Thomas DS2-1-38*, DS2-1-80, DS2-2-144 DS2-2-77, OS-32* Nahnsen, Sven DS2-1-14* Miller, Samantha DS2-2-23, DS2-4-22, Naito, Yasuhiro DS2-2-101, DS2-2-24 DS2-4-42 Nakahigashi, Kenji OS-23 Miller-Jensen, Kathryn DS3-1-07 Nakajima, Akihiko DS2-1-92* Ming-Tat, Ko DS3-3-20 Nakamura, Takeshi DS2-1-54 Minton, Nigel P. DS2-1-107 Nakayama, Shin-ichi DS3-3-38 Mir, Saqib DS1-2-05 Naldi, Aurélien DS2-1-110, DS2-2-140 Mironov, Vladimir A-08 Nandy, Subir Kumar DS2-1-30* Mirschel, Sebastian DS3-3-16, DS3-3-17* Naoki, Honda DS2-2-59* Mirzoeva, Olga DS2-1-70 Nardelli, Maria DS2-4-14 Miyano, Satoru DS2-1-119, DS2-1-50 Nartsissov, Yaroslav DS2-2-146 Mjolsness, Eric A-37 Naruo, Yoshimi DS2-1-13* Mobini, Reza DS2-3-08* Natter, Klaus DS2-4-31* Moellering, Robert DS2-4-05 Naumann, Michael DS2-1-115 Mohamed Suleiman, Khalifa DS1-4-28* Naumov, Andrey DS1-1-67, DS2-1-96 Mol, Xavier DS2-2-136 Nedelec, Francois P-16* Molenaar, Douwe A-06, DS2-4-27 Nees, Matthias DS3-3-32 Molenaar, Piet A-07* Nehaniv, Chrytopher L. DS2-2-103 Molina, Franck A-30, DS2-2-65, DS3-2-10 Nelander, Sven DS2-3-18*, DS2-3-18 Moné, Martijn J. DS2-1-73*, DS2-1-93, Nelson, David DS2-1-02 DS2-1-81, P-15 Nerman, Olle A-19, OS-13 Monnier, Annabelle DS1-4-27 Neufeld, Zoltan DS2-1-120* Monostory, Katalin DS1-1-44 Neumann, Leo DS2-1-105, DS3-4-04 Montañez, Raúl DS1-1-12 Neves, Ricardo DS3-1-17*

Montelius, Andreas OS-49 Nguyen, Nam-Phuong D. A-45, DS3-3-02 Index Author Monuki, Edwin DS2-1-91 Nicholson, Jeremy DS2-3-09 Moodie, Stuart DS1-3-24 Nici, Marco DS2-1-21 Moon, Simon DS2-1-41 Nickel, Peter J. DS2-1-23* Moore, Travis DS2-1-79* Nicolay, Klaas DS1-1-06, DS1-1-48 Mootha, Vamsi DS1-1-26 Nie, Qing DS2-1-79, DS2-1-89 Morant, Pierre-Emmanuel DS1-4-27, DS2-2-153 Niebel, Anja DS1-3-03 Moraru, Ion A-38*, DS1-2-17*, DS1-2-18*, Niederalt, Christoph DS1-3-04 DS3-3-05* Nielsen, F DS2-1-34 Moray, Campbell DS2-1-123 Nielsen, Jens A-16*, A-18, A-20, DS1-1-07, Morcos, Faruck DS3-3-37* DS1-1-10, DS1-1-20, Morgan, Jeffrey DS2-2-54 DS1-1-24, DS1-1-36, Mori, Hirotada DS2-4-13, OS-23* DS1-1-40, DS1-1-42, Morillas, Montse DS2-1-03 DS1-1-43, DS1-3-19, Moriya, Hisao DS1-3-12, DS3-4-22, OS-11 DS2-1-30, DS2-1-47, Morrison, Jonathan DS3-4-01 DS2-4-26, DS3-2-16, OS-24, Morrissey, Edward DS2-2-114 P-20* Mosekilde, Erik DS2-2-86, DS2-3-13 Nielsen, Michael N. DS1-1-20 Motta, Jean-Paul DS1-4-27 Nielsen, Poul DS1-2-09, DS1-2-14 Moura, Alessandro DS2-4-22, DS2-4-23 Nielsen, U.B. P-19 Mourao, Andre DS1-1-47 Nierop, Andreas DS2-3-10 Mozga, Ivars DS1-1-64*, OS-43 Nikkilä, Janne DS1-1-27 Mpindi, John-Patrick DS3-3-32 Niklas, Jens DS1-3-09 Msadek, Tarek A-30 Nikoloski, Zoran DS1-4-14* Mudhar, Ramandeep DS1-1-28, DS2-4-06 Nilsson, Bengt OS-13 Mueller, Margareta DS2-1-21 Nilsson, Ola OS-13 Mueller, Stefan DS2-2-26* Nilsson, Roland DS1-1-26*, DS2-3-07 Muench, Richard DS2-4-32 Nishino, Taiko DS2-2-15* Muggleton, Stephen DS2-2-133 Noack, Stephan DS2-2-43 Mukhopadhyay, Ranjan DS2-4-16 Nobata, Chikashi A-13* Mullen, Peter DS1-3-24 Nock, R DS2-1-34 Müller, Stephanie DS2-1-23 Nogami, Satoru DS2-4-02, OS-36 Mullins, JJ DS2-1-55 Noh, Tae-Hwan DS2-4-40, DS2-4-41 Munoz-Garcia, Javier DS2-1-120 Nöh, Katharina DS2-2-43* Munteanu, Andreea DS2-2-92* Noirot, Philippe A-30 Mura, Ivan DS2-2-135* Nookaew, Intawat DS1-1-36* Murata, Shinya OS-08* Noor, Fozia DS1-3-09 Murugan, Prem Kumar DS2-1-78* Noori, Peri DS2-3-02, DS2-3-07 Muruganujan, Anushya DS3-3-12 Norbeck, Joakim DS2-1-49* Musters, Mark DS2-2-07*, DS2-4-30* Nordheim, Alfred DS2-1-14

ICSB 2008 223 Nordlander, Bodil A-24, DS2-1-111, DS2-1-118, Papadakis, Manos A. DS1-1-43* DS2-1-03 Papin, Jason DS1-1-15, DS2-4-17 Nordlander, Carola OS-22 Papp, Balazs DS1-1-05*, DS1-1-39, Nordling, Torbjörn E.M. DS3-4-02*, DS1-1-22 DS1-1-45 Norrheim Larsen, Laila DS3-3-62 Parent, Benjamin DS2-2-153 Nörtemann, Bernd DS2-2-45 Park, Inju DS1-1-62* Notebaart, Richard DS1-1-45*, DS2-4-27 Park, Taesung OS-12 Nour-Mohammadi, Ghorban DS1-4-26 Park, Young-Jin DS2-4-40, DS2-4-41* Novák, Béla A-35, P-14*, OS-02 Parmar, Jignesh DS2-1-106* Nyström, Thomas DS2-2-112 Parshina, Evgeniya DS2-3-13 Pascussi, Jean-Marc DS1-1-44 O Paszek, Pawel DS2-1-02 Paton, Norman A-04, DS1-1-65, DS1-2-15 Oberthuer, Angela A-42 Paul, Perrine DS3-3-15* Obolenskaya, Maria DS3-3-31 Paulsson, Johan DS3-1-06* Oda, Kanae DS3-3-52 Paw, Barry DS1-1-26 Oestergaard, Mikkel DS3-4-01 Pawson, Tony P-05 Ogawa, Yukino DS3-4-18* Pearce, Roger S. DS1-4-28, DS1-4-28 Oh, Eulsik DS2-2-151 Peccoud, Jean DS1-2-03* Ohnuki, Shinsuke DS2-4-02, OS-36* Peddinti, Gopalacharyulu OS-06* Ohta, Jun DS1-1-70* Pedersen, Anders Gorm OS-07 Ohta, Nobuyuki DS3-3-39* Pedro, Mendes A-51 Ohya, Yoshikazu DS2-4-02*, OS-36 Peeters, Ralf DS2-2-75 Oiwa, Kazuhiro DS2-2-141 Pel, Herman DS2-4-38 Ojala, Kalle DS3-3-32 Pelet, Serge DS2-1-03, DS3-1-03* Oka, Satomi DS2-4-02 Pelkmans, Lucas DS2-1-45 Okada, Chihiro A-52 Pena-Castillo, Lourde DS2-4-03 Okazaki, Naoaki A-13 Penalva, Luiz O. DS2-1-11* Okeguchi, Kazuyuki DS2-1-50 Penkler, Gerald A-27, DS1-1-34*, DS3-3-07 Olivares, Roberto DS1-1-40* Penttilä, Merja DS1-1-68 Oliver, Ratmann DS2-2-04 Perala, Merja DS3-3-32 Oliver, Stephen G. A-04, DS1-1-31, DS1-1-05, Peres, Sabine DS2-2-65*

Author DS1-1-32, DS1-1-52, Periyasamy, Sathish DS2-2-118* Index DS1-1-65, DS2-4-15, OS-38 Perkins, Andy D. DS2-3-08 Olivier, Brett G A-27, DS3-3-23*, DS3-3-44 Perkins, Edward DS3-4-06, DS3-4-13 Olle, Nerman OS-40 Persson, Ronnie DS3-3-56 Olsen, Lars Folke DS1-1-51*, DS1-1-57 Pesce, Gustavo DS3-1-05*, DS2-4-04 Olsen, Peter DS1-1-42 Pesch, Martina DS1-4-03, DS1-4-23 Olsson, Lisbeth DS1-1-60, DS2-4-26, OS-24 Pescini, Dario DS2-1-112 Olteanu, Violeta DS2-4-33 Petelenz, Ela DS2-1-03 O’Malley, Brendan DS1-1-08* Petelenz, Elzbieta DS2-1-118*, DS3-1-12 Omholt, Stig W. DS2-2-117, DS2-2-137, Peter, Matthias DS2-1-03, DS3-1-03 DS2-2-67, DS2-4-45, P-23*, Petranovic, Dina DS1-1-07 DS2-1-58, DS2-2-120 Petrovic, Uros DS2-1-53* Onami, Shuichi A-32* Pettifer, Steve A-51, DS3-3-49 Onsan, Z. Ilsen DS2-1-47 Pfeifer, Andrea C. DS2-1-101* Oren, Moshe DS2-1-04 Pfeiffer, Thomas DS1-1-25 Orešič, Matej DS1-1-27, DS2-3-15, A-49*, Pforr, Carina DS2-1-105, DS3-4-04 DS1-1-02*, OS-06 Pharoah, Paul DS3-4-01 Orrell, David DS1-3-05* Pi, Haiwei DS3-4-11 Orr-Urtreger, Avi DS2-3-05* Pickersgill, Laura DS1-1-08 Østby, Ivar DS2-2-117*, DS2-2-137 Picotti, Paola DS1-1-38 Ostheimer, Gerry P-05 Pierce, Eric DS1-1-26 Otero, Jose Manuel DS2-4-26*, OS-24, DS1-1-43 Pierre, Philippe DS2-1-31 O’Toole, Paul DS3-3-47 Pillay, Che DS1-1-19 Ottosson, Lars-Göran DS2-1-111* Pilpel, Yitzhak DS2-1-04, DS2-2-124 Oyama, Masaaki DS2-1-116*, DS2-1-117, Pincus, David DS2-4-04 DS2-1-119, DS2-1-27 Pir, Pinar DS1-1-31, DS1-1-52 Øyehaug, Leiv DS2-2-117, DS2-2-137*, Piras, Vincent DS2-1-102*, DS2-2-87* DS2-1-58 Pisarev, Andrei DS3-3-51 Ozbabacan, Ece DS2-1-82* Pisto, Tommi DS3-3-32 Ozyamak, Ertan DS2-4-23 Plahte, Erik DS2-2-120, DS2-2-67 Plotkin, Joshua DS3-2-04 P Podlesnaja, Svetlana DS2-3-20 Pogorelko, Gennady DS1-4-24* Pace, E. A. P-19 Politi, Antonio P-15 Pachkov, Mikhail DS2-1-124 Pollard, Chris DS1-3-27 Padiadpu, Jyothi DS2-2-110 Polouliakh, Natalia DS2-1-34* Painter, Kevin J. DS1-4-12 Poltz, Rainer DS2-1-115* Pal, Csaba DS1-1-05 Ponder, Bruce DS3-4-01 Palmisano, Alida DS2-2-135 Pontén, H. DS1-4-21 Palsson, Bernhard O. DS2-2-25, DS3-4-05 Poole, Robert DS2-2-89 Panagiotou, Gianni DS1-1-43 Poolman, Mark DS1-4-07, DS2-2-122, Panovska, Jasmina DS1-1-08 DS2-4-07 Papachristodoulou, Antonis DS2-2-18, DS3-4-16 Posas, Francesc DS2-1-03

224 ICSB 2008 Posdziech, Florian DS2-1-101 Reyes-Palomares, Armando DS1-1-12* Pospisil, Heike DS2-2-50 Rezen, Tadeja DS1-1-44 Postnov, Dmitry DS2-2-86 Rho, Seong-Hwan DS2-1-14, DS2-4-28* Potapov, Anatolij DS2-2-83* Rialle, Stephanie DS3-2-10* Potselueva, Margarita DS1-1-67, DS2-1-96 Richard, Morgiane DS2-4-22*, DS2-4-42 Poulsen, Allan Korsgaard DS1-1-51 Richardson, Sylvia DS2-2-04 Prakash, Nilima DS2-2-11 Rink, Jochen DS2-1-98 Prätzel-Wolters, Dieter DS2-2-63 Rintala, Eija DS1-1-68 Prehn, Jochen DS2-1-108 Rippe, Karsten DS2-1-72 Pronk, Jack A-06 Ritter, Daniel DS2-2-60 Pronk, Jack T DS1-1-59 Rivet, Catherine DS3-4-17* Protas, Oleksiy DS3-3-31 Robert, Martin A-15*, DS1-1-49* Prusinkiewicz, Przemyslaw DS1-4-02*, DS1-4-33, Roberts, Mark DS3-4-16* DS3-3-57 Roberts, Seth OS-33 Przemeck, Gerhard DS2-1-88 Robinson, Sarah DS1-4-33* Przulj, Natasa DS2-2-116*, DS2-2-39*, Robson, Paul DS3-1-13 DS2-2-41*, DS3-3-11* Roca, Josep A-54*, DS2-3-21 Pulkkinen, Otto DS3-1-10* Rocha, Isabel DS2-2-94 Puzanov, Mikhail DS2-2-149 Rocha, Miguel DS2-2-94 Rockwell, Daniel DS3-1-05 Q Rodnenkov, Oleg DS2-3-13 Rodríguez, Diego DS2-3-21 Qannari, Mostafa DS3-3-62 Rodriguez Martinez, Maria DS2-2-124* Qian, Hong DS2-2-113 Rodriguez-Caso, Carlos DS2-2-64* Quashie, Peter DS1-1-28, DS2-4-06 Rodriguez-Fernandez, Maria DS2-2-71*, DS2-2-78* Quast, Karsten DS1-3-26 Roeder, Adrienne A-37 Querin, Lorenzo DS2-1-59 Rogon, M.Z. DS2-1-40*, DS2-2-13 Rohwer, Johann A-27, DS1-1-19, DS3-3-23, R DS3-3-35, DS3-3-44, DS1-1-11 R. Banga, Julio DS2-2-71, DS2-2-78, Rojas, Isabel DS1-2-05 DS3-3-48 Roland, Siezen DS1-1-45 Rädler, Joachim DS3-1-14 Rolfe, Matt DS2-2-89, DS3-3-25 Radulescu, Ovidiu DS2-2-106 Roling, Wilfred DS2-4-10* Radvanyi, François DS2-1-57 Romanel, Alessandro DS2-2-135

Rai, Navneet DS3-2-19* Romano, M. Carmen DS2-2-23, DS2-2-81* Index Author Raineri, Emanuele DS1-1-47, DS3-2-02 Ronneberger, Olaf DS1-4-15 Rak, Sabina DS2-3-08 Roos, Christophe OS-43 Raman, Karthik DS1-3-25 Rosa, Agostinho DS2-2-51 Ramialison, Mirana DS2-1-28* Rosfors, Stefan DS2-3-07 Ramirez, Fidel DS1-2-08* Rossi, Derrick DS3-4-24 Ramkumar, Krishna DS3-2-19 Rosu, Haret DS2-2-76 Ramlal, Nishant DS1-1-28, DS2-4-06 Roth, Frederick DS3-4-01* Ramsell, Laura DS1-3-05 Rother, Kristian DS1-1-35* Rand, David DS2-1-02, DS2-1-55 Roubos, Hans DS2-4-38 Rantala, Juha DS3-3-32 Roumani, Ali DS2-4-13 Rantanen, Ari DS1-1-66* Rousu, Juho DS1-1-66 Rasajski, Marija DS2-2-41 Rowland, Jem DS3-4-08 Rash, Bharat DS1-1-31, DS1-1-32, Rozman, Damjana DS1-1-44* DS1-1-52 Rudolf, Fabian DS3-1-03 Rashidi, Armin DS2-2-49* Ruebenacker, Oliver DS1-2-18 Rateitschak, Katja DS2-1-109 Ruenwai, Rawisara DS1-1-07* Rausenberger, Julia DS1-4-11 Rundell, Ann DS3-3-21 Rawlings, Christopher DS2-2-133 Ruoff, Peter DS1-1-63, DS2-4-14 Rea, Brian A-13 Ruohonen, Laura DS1-1-68 Redowicz, Jolanta DS2-1-94 Ruppin, Eytan DS2-4-39 Reed, Michelle DS2-3-21 Russo, Giovanni DS3-2-08* Regina, Samaga DS2-1-06 Rutkis, Reinis DS1-1-14 Rehberg, Markus DS2-2-71 Ruttenberg, Alan DS1-2-02* Rehm, Markus DS2-1-108* Ryan, Sheila DS2-1-02 Reich, Jens Georg DS1-1-58 Ryazanova, Ludmila DS2-2-86 Reichart, Thomas DS2-1-78 Rybakova, Katja N DS2-1-81* Reifman, Jaques DS2-2-70* Reinitz, John DS2-2-143, DS3-3-51, OS-16 S Reittie, Joyce DS3-1-17 Remm, Maido OS-43 Sacco, Elena DS2-1-71 Rempel, Michael DS3-3-16 Saebo, Solve DS2-2-72 Repsilber, Dirk OS-18* Sætre, Rune DS3-3-52* Retter, Ida DS2-4-32 Saez-Rodriguez, Julio DS2-1-06*, DS2-1-42, Reuss, Matthias A-41*, DS1-1-21, DS1-1-63, DS2-1-60, DS2-2-69, DS1-3-03, DS2-1-78, DS3-3-17 DS2-2-31, DS2-4-09, Sahlberg, Niko DS3-3-32 DS2-4-14 Sahle, Sven DS2-2-28*, DS3-3-03 Revelles, Olga DS3-1-19* Sahlin, Patrik A-37, DS1-4-06, DS1-4-30* Reyero Vinas, Natàlia G. DS3-4-06 Sahoo, Debashis DS3-4-24

ICSB 2008 225 Saithong, Treenut DS1-4-12* Schroeter, Anja DS2-4-39* Saito, Ayumu DS2-1-50 Schuerrle, Karsten A-01* Saito, Natsumi DS1-1-49 Schuetz, Robert DS1-1-46* Saito, Rintaro OS-08 Schultz, Iman DS1-1-26 Sakumura, Yuichi DS2-1-54 Schuster, Stefan DS1-1-25*, DS2-4-39 Sakurada, Takeshi A-52 Schwartz, Christian DS1-4-27 Salama, Rafik DS2-2-74* Schwartz, Jean-Marc DS1-3-13* Salazar, Carlos DS2-1-20, DS2-1-86 Schwarz, Ulrich DS2-2-37 Salazar, Margarita DS3-2-16* Schwarzwälder, Eva DS2-2-90 Samaga, Daniel DS2-2-61* Schwikowski, Benno A-30 Samaga, Regina DS2-2-69*, DS3-3-18 Scott, Matthew DS2-4-19* Samalikova, Maria DS2-1-66 Sealfon, Stuart DS2-1-62 Samnegård, Ann DS2-3-07 Sebastiano, Battaglia DS2-1-123* Samsonov, Alexander DS2-2-143 Sedwards, Sean DS2-2-135, DS2-2-142* Samsonova, Maria DS2-2-143, DS3-3-51, OS-16 See, Violaine DS2-1-02 San-Bento, Rita DS1-3-19 Segal, Eran P-06* Sanchez-Diaz, Patricia DS2-1-11 Segre, Daniel DS2-2-33, DS2-2-35 Sánchez-Jiménez, Francisca DS1-1-12 Seita, Jun DS3-4-24* Sander, Chris P-01* Selbig, Joachim DS1-4-14, OS-18 Sanders, Phil DS3-2-02 Selivanov, Vitaly A-54, DS1-1-69, DS2-3-21 Sano, Hitomi DS2-2-101, DS2-2-24* Selvarajoo, Kumar DS2-1-102, DS2-1-32, Sanoudou, Despina A-02* DS2-1-90, DS2-2-87 Sansone, Susanna-Assunta DS1-2-06* Semba, Kentaro DS2-1-117, DS2-1-119 Sara, Henri DS3-3-32 Semisalov, Boris DS2-2-149 Sasaki, Yutaka A-13 Semple, Jennifer OS-48 Sasidharan, Kalesh DS2-2-87 Semprini, S DS2-1-55 Sassi, Holly DS2-4-03 Sendhoff, Bernhard DS2-2-147 Sato, Masayuki J DS3-4-23 Sendín, José-Oscar H. OS-26* Sattler, Michael DS1-1-47 Seppänen-Laakso, Tuulikki DS1-1-27, DS2-3-15 Sauer, Uwe A-30*, DS1-1-03, DS1-1-38, Sepulchre, Jacques-Alexandre DS2-1-22*, OS-25 DS1-1-46, DS1-1-66, Serra, Eduard DS2-4-04 DS3-1-19, P-13* Serrano, Luis DS1-1-47, DS2-1-56,

Author Sauro, Herbert M. DS2-2-113, DS3-3-40 DS3-2-02, P-21* Index Sauter, Thomas DS1-1-04, DS2-2-90, Shabala, Sergey DS2-2-129 DS3-4-12 Shabala, Svetlana DS2-2-129 Sautot, Caroline A-30 Shabi, Uri DS3-2-03*, P-22 Saviranta, Petri DS3-3-32 Shadrin, Aleksey DS3-3-60 Sawodny, Oliver DS1-1-04, DS2-2-90, Shah, Priya DS3-1-07 DS3-4-12 Shalgi, Reut DS2-1-04* Sazuka, Naoya DS2-2-144, DS2-2-145* Shamsi, Keyvan DS1-4-26* Scalcinati, Gionata OS-24* Shanley, Daryl A-46*, DS2-2-49 Schaber, Joerg DS2-1-03*, DS1-1-56, DS2- Shapiro, Bruce A-37 1-118, DS2-1-95, DS2-2-01, Shapiro, Ehud P-22* DS3-3-46 Sharifpoor, Sara DS2-4-03 Schaefer, Eberhardt DS1-4-11, DS1-4-25 Sharipov, Ruslan DS1-1-67*, DS1-2-19*, Schaff, James A-38, DS1-2-17, DS3-3-05 DS2-1-122*, DS2-1-125, Schaffer, David DS3-1-07 DS2-1-96, DS2-2-149* Scheja, Ludger DS2-2-50 Sharkey, Kieran DS2-4-14 Schellenberger, Jan DS3-4-05* Sharma, Sapna DS1-4-20 Schellmann, Swen DS1-4-03, DS1-4-23 Shatalin, Yuriy DS1-1-67, DS2-1-96* Scherzinger, Tobias DS2-1-21 Shaw, Stanley DS2-3-06* Schilling, Marcel DS2-1-17*, DS2-1-20, She, Bin DS2-1-86* DS2-1-26, DS2-1-33, Shelton, Rebecca DS1-2-03 DS2-1-86 Shieh, Grace Shwu-Rong DS2-2-38* Schilstra, Maria J. A-14, DS2-2-103 Shiina, Marisa DS3-4-19 Schittek, Birgit DS2-1-14 Shimamura, Teppei DS2-1-50 Schlatter, Rebekka DS3-4-12* Shimayoshi, Takao DS3-3-30* Schloeder, Johannes DS3-4-07 Shin, Dongkwan DS2-2-131* Schmid, Ramona DS1-3-26* Shin, Kyung Sook OS-45 Schmidt, Esther DS1-2-04* Shin, Sung-Young DS2-1-48* Schmidt, Henning DS1-2-11*, DS2-1-109, Shin, Yong-Jun OS-27* DS2-1-64*, DS2-2-40*, Shiraishi, Tetsuya DS2-2-144, DS2-2-145 DS2-2-63, DS3-3-09*, Shoemaker, Jason E. DS3-4-06* DS3-3-28* Shojaie, Sharareh DS1-4-31, DS1-4-32 Schmitz, Joep DS1-1-06, DS1-1-48 Siegmund, Hans-Ulrich DS1-3-04 Schmitz, Ulf OS-43 Siemann-Herzberg, Martin DS1-1-21 Schneider, Konstantin DS2-4-21* Siewers, Verena DS1-3-19* Schneltzer, Elida DS2-1-88 Sikora, Marcin DS3-3-37 Schoeberl, Birgit P-19* Silk, David DS2-3-09 Schoen, Verena DS2-4-28 Sillitoe, Kate DS2-1-02 Schoettler, Anja DS2-2-50* Silveira, Angela DS2-3-07 Scholz, Matthias A-55* Simell, Olli DS1-1-27 Schomburg, Dietmar DS3-3-59 Simeonidis, Evangelos A-04, DS1-1-30*, DS1-1-55, Schönenberger, Philipp DS2-4-12 DS1-1-65 Schreiber, Stuart DS2-3-06 Simons, Veronika A-48

226 ICSB 2008 SITCON, Consortium DS2-1-69 Stevens, Robert A-08 Sjöberg, Paul DS2-2-79* Stilgenbauer, Stephan DS2-1-72 Sjövall, Peter OS-17 Stitt, Mark DS1-4-01* Skanda, Dominik DS2-2-48 Stoll, Gautier DS2-1-69* Skogsberg, Josefin DS2-3-02, DS2-3-07 Stoof, Cor A-27, DS3-3-07 Slaby, Oliver DS2-2-47 Stratford, Kevin DS1-4-12 Slepchenko, Boris A-38 Strömberg, Anna OS-49* Sletta, Håvard DS2-4-34 Studholme, David OS-20 Sliwa, Piotr DS2-1-103 Stuelke, Joerg DS3-3-33 Smallbone, Kieran A-04, DS1-1-30, DS1-1-55*, Stumpf, Michael DS2-2-04 DS1-1-65 Stys, D. OS-50* Smedsgaard, Jorn DS1-1-10 Su, Wen-Hui DS2-3-04* Smid, Eddy DS2-4-27 Subbanna, Nagesh DS2-2-110* Smith, Colin P DS3-3-55 Subramanian, Aravind DS2-3-06 Smith, David A-35 Suematsu, Makoto DS2-2-15 Smith, Paul DS1-3-06 Sugano, Sumio DS2-1-117, DS2-1-119 Snitkin, Evan DS2-2-35* Sugimura, Haruka DS3-3-45 Snoep, Jacky A-27*, DS1-1-01*, DS1-1-19, Sukhomlin, Tatyana DS1-1-67, DS2-1-96 DS1-1-34, DS3-3-07, Sunnåker, Mikael DS2-2-08, DS3-3-43* DS3-3-27, OS-02, DS1-1-11, Surkova, Svetlana DS2-2-143, OS-16 DS1-3-01, DS2-4-14 Surovtsev, Ivan DS2-2-54 Snyder, Michael P-08* Suzuki, Ken DS2-1-77* Sobhe Bidari, Pooya DS1-4-31, DS1-4-32 Svedhem, Sofia DS1-3-21, OS-17* Soebiyanto, Radina P. DS2-1-38 Svensson, P. DS1-4-21 Soga, Tomoyoshi DS1-1-49, DS3-4-18 Swainston, Neil A-04, DS1-1-65, DS1-2-15, Soldatova, Larisa DS3-4-08 DS3-3-61* Solé, Ricard V. DS2-2-92 Sweede, Matthew A. DS1-2-03 Solmaz, Erim DS1-4-04* Sweetlove, Lee DS1-4-07 Somersalo, Erkki DS3-3-53 Swinton, Jonathan DS1-3-27 Soneson, Charlotte DS2-3-22* Sydorovych, Inna DS2-2-20* Sonesson, C. DS1-4-21 Sysi-Aho, Marko DS1-1-27 Song, Eun-Sung DS2-4-40, DS2-4-41 SysMO, SUMO Consortium A-28 Sonnhammer, Erik DS3-4-14* Szabowski, Axel DS2-1-72, DS2-2-13, Sonntag, Sebastian DS1-4-25* DS2-1-40 Soranzo, Nicola DS1-1-41, DS2-2-53* Szewczyk, Adam OS-37

Soreq, Hermiona A-02 Index Author Sorger, Peter DS2-1-06, DS2-1-42, T DS2-1-51, DS2-1-60, DS3-1-09, DS3-4-03 Tabaka, Marcin DS3-1-18* Sorger, Peter K DS2-2-69 Tachikawa, Masashi DS2-2-06* Soriano, Jordi DS2-2-124 Tacutu, Robi DS2-2-34* Sorokin, Anatoly DS1-4-05 Tadros, Menerva DS3-1-15 Sosnovtseva, Olga DS2-2-86, DS2-3-13 Takagi, A DS2-1-34 Sotiropoulos, Vassilios DS3-2-07 Takagi, Hiroaki DS3-4-23* Sott, Kristin DS3-1-12 Takahashi, Hiro DS2-3-19* Southan, Christopher DS1-3-14* Takahashi, Koichi A-52 Sparkes, Andrew DS3-4-08 Takalo, Jouni DS2-1-61 Spasic, Irena A-04, DS1-1-30, DS1-1-65 Takeda, Sachiko DS2-1-16 Spencer, Sabrina DS2-1-51, DS3-1-09* Takeuchi, Maria DS2-2-101* Spiesser, Thomas DS2-2-130* Takeuchi, Osamu DS2-1-85 Spiller, David DS1-2-15, DS2-1-02 Takeuchi, Rikiya OS-23 Sprenger, Georg DS2-4-09 Takolander, Rabbe DS2-3-07 Srayko, Martin P-16 Tamaddoni-Nezhad, Alireza DS2-2-133 Sreenivasan, Varun DS3-2-19 Tamas, Markus DS1-1-56 Srenath, Sree N. DS2-1-38 Tamminen, Anu DS1-1-68 Sriniva, Bhylahalli Purushott DS1-4-03, DS1-4-23 Tanaka, Hiromasa DS2-1-113* Srinivasan, Shyam DS2-1-91* Tanase-Nicola, Sorin DS3-1-04 Stagljar, Igor DS2-4-01* Taniguchi, Yuichi DS3-1-08 Stalidzans, Egils DS1-1-64, OS-43* Tanticharoen, Morakot DS1-1-07 Stansfield, Ian DS2-2-81 Tasaki, Shinya DS2-1-117, DS2-1-119* Stansvik, A. DS1-4-21 Tasan, Murat DS3-4-01 Stark, Jaroslav DS2-1-19, DS2-1-41 Tavi, Pasi DS2-1-61 Starmans, Maud DS2-1-41 Tegnér, Jesper DS2-2-16, DS2-3-02*, Stebel, Sabine A-33 DS2-3-07, OS-03 Stefan, Melanie I DS2-1-12* Tei, Hajime DS3-4-18 Steijaert, Marvin DS2-2-56* Teixeira, José DS2-2-51 Stein, Lincoln DS1-2-04 Tekir, Saliha DS2-1-82 Steiner, Till DS2-2-147* Telle-Hansen, Vibeke H. DS3-3-62 Steinmetz, Katrin DS3-3-16*, DS3-3-18* Ten Eikelder, Huub M.M. OS-09, DS2-2-56 Steinmetz, Lars P-26 Ten Wolde, Pieter Rein DS3-1-04* Stelling, Joerg A-30, DS2-2-09, DS2-2-29, Tenazinha, Nuno DS2-2-115* DS3-2-01* Terzer, Marco DS2-2-29* Stergiou, Lilli DS2-1-45* Sternberg, Michael DS2-2-133 Teusink, Bas A-06*, DS1-1-16*, DS1-1-45, Steuer, Ralf DS2-2-98*, DS2-4-14 DS2-4-27*, DS2-4-30

ICSB 2008 227 Thalhammer, Armin DS2-4-31 U Thammarongtham, Chinae DS1-1-36 Thattai, Mukund DS3-2-19, OS-44 Ude, Susanne DS2-4-28 The AMPKIN consortium, Ueberham, Elke DS2-1-23 EC-funded A-23 Ueda, Masahiro DS2-2-36, DS3-4-23 The CELLCOMPUT Consortium, Ueno, Kazuko DS2-1-50 EC-Funded A-24 Uetzmann, Lena DS2-2-109 The SYSTEMSBIOLOGY Uhlendorf, Jannis DS3-3-24 Consortium, EC-funded A-19 Uhlmann, Frank P-14 The UNICELLSYS Consortium, Ukkonen, Esko DS1-1-66 EC-funded A-17 Ulgen, Kutlu O. DS2-1-82 The YSBN Consortium, Ullrich, Alexander DS1-1-29* EC-funded A-16 Ulven, Stine Marie DS3-3-62 Theis, Fabian DS2-1-88*, DS2-1-99, Urban, J. OS-50 DS2-2-109, DS2-2-11 Urdiales, José Luis DS1-1-12 Thieffry, Denis DS2-1-110, DS2-2-140 Usadel, Björn DS1-4-18 Thiel, Marco DS1-1-13, DS2-2-23*, Usaite, Renata DS1-1-40 DS2-2-81 Uscatescu, Victor DS1-1-28, DS2-4-06 Thiele, Ines DS2-2-25* Ushioda, Junya DS2-1-114 Thierry, Alain DS3-2-10 Thiesen, Hans-Juergen DS2-2-68 V Thingnes, Josef DS2-1-58* Thodima, Venkata DS3-4-13 Vacun, Gabriele DS1-3-03 Thomas, Paul DS3-3-12 Vafiadaki, Elizabeth A-02 Thommen, Quentin DS1-4-27, DS2-2-153 Vainio, Paula DS3-3-32 Tian, Weidong DS3-4-01 Valachovic, Martin DS2-4-14 Tiedemann, Hendrik DS2-1-88 Valente, André X. C. N. DS3-3-13*, OS-33* Tiemann, Christian DS2-1-41 Valla, Svein DS2-4-29* Timmer, Jens DS1-4-03, DS1-4-11, Vallon, Tobias DS2-4-09 DS1-4-15, DS1-4-23, van den Bosch, Paul DS2-2-07 DS1-4-25, DS2-1-101, van den Brink, Joost DS1-1-59* DS2-1-17, DS2-1-20, Van Dijl, Jan Maarten A-30

Author DS2-1-23, DS2-1-26, Van Driel, Roel P-15*, A-40*, DS2-2-121 Index DS2-1-33, DS2-1-99, van Dyk, Dewald DS2-4-03 DS2-2-02*, DS2-4-28, van Eunen, Karen DS1-1-61* DS3-3-06, DS3-4-21 van Gend, Carel DS1-1-01, DS3-3-07, Timmers, Mieke OS-05 DS3-3-27*, A-27 Timr, S. OS-50 van Leeuwen, Johannes P.T.M. DS2-1-73 Tindall, Marcus DS1-1-08 van Nimwegen, Erik DS2-1-124 Titorenko, Vladimir DS1-1-28*, DS2-4-06* van Noort, Vera DS1-2-16 Tlusty, Tsvi DS2-2-124 van Nuland, Rick DS1-3-08 Tobor-Kaplon, Maria DS2-4-10 van Riel, Natal DS1-1-06*, DS1-1-48*, Togashi, Yuichi DS2-2-36* DS2-1-15*, DS2-1-41*, Toivari, Mervi DS1-1-68 DS2-2-07 Tokovenko, Bogdan DS3-3-31* van Tuijl, Arjen DS1-3-08 Toksoy Oner, Ebru DS2-4-18 Van Vleet, Jennifer R. H. OS-24 Tollervey, David DS3-2-04 Vanags, Juris OS-43 Tolstykh, Nikita DS1-2-19, DS3-3-60 Vandemoere, Constant DS1-4-27, DS2-2-153 Toma, Malda Maija DS1-1-14 Vanek, J. OS-50 Tomassini, Marco DS2-2-46 Vanoni, Marco DS2-1-59, DS2-1-63*, Tomita, Masaru A-52, DS1-1-49, DS2-1-102, DS2-1-66, DS2-1-71*, DS2-1-32, DS2-1-90, DS2-2-95, DS2-2-96 DS2-2-101, DS2-2-15, Vastrik, Imre DS1-2-04 DS2-2-24, DS2-2-87, Vavouri, Tanya OS-48* DS3-4-18, OS-21 Veflingstad, Siren R. DS2-2-67 Tomshine, Jonathan DS3-2-07 Velagapudi, Vidya OS-06 Torres, Luis DS2-2-76* Velarde, Giles DS3-3-55 Traas, Jan DS1-4-06 Vemuri, Goutham DS1-1-20*, DS1-1-60 Trakhinin, Yuriy DS2-2-149 Venkatesh, K.V. DS3-2-19, DS2-1-106 Trané, Camilla DS2-1-126* Ventura, Alejandra C DS2-1-22 Tripodi, Farida DS2-1-66 Vera Gonzales, Julio A-25 Truembach, Dietrich DS2-2-11 Vergauwen, Lucia OS-05 Tsiarentsyeva, Viktoryia DS2-1-66 Verma, Malkhey DS2-4-14 Tsuchiya, Masa DS2-1-102, DS2-1-32, Vermeulen, Wim P-15 DS2-1-90, DS2-2-87 Veyrieras, Jean-Baptiste P-26 Tsujii, Junichi DS3-3-52, A-13, A-43 Vicente, António DS2-2-51 Tsukada, Yuki DS2-1-54* Vidal Puig, Antonio DS2-3-15, DS1-1-02 Tsuru, Saburo DS2-1-05, DS2-1-114* Vie, Ane Kjersti DS2-4-34 Tsuruoka, Yoshimasa A-13, A-43 Vielhauer, Oliver DS2-4-09 Tuohy, Kieran DS2-3-09 Villà, Jordi A-54 Turan, Nil A-54 Villeger, Alice DS3-3-49* Turner, Frances DS2-2-133 Villesen, Palle DS2-3-03* Tuschl, Thomas DS2-1-76 Vinga, Susana DS2-2-115 Tyrer, Jonathan DS3-4-01 Vinnicombe, Glenn DS3-1-06 Vishveshwara, Saraswathi OS-42

228 ICSB 2008 Vizán, Pedro DS1-1-69 Willy, Paul DS3-3-03 Vogel, Christine DS2-1-11 Windram, Oliver P. F. DS1-4-20 von Bornstaedt, Gesa DS2-2-121*, P-15 Wingender, Edgar DS2-2-83, A-02 von Gruenberg, Hans-Hennig DS2-4-31 Wingreen, Ned DS2-4-16 von Kamp, Axel DS3-3-54* Winkler, David DS2-2-100, DS2-2-127 von Kleist, Max DS1-3-07* Winter, Dominic DS2-1-17 von Zglinicki, Thomas A-46 Wipat, Anil A-46 Vonder Mühll, Daniel A-39* Wishart, Jill DS1-1-65 Vongsangnak, Wanwipa DS1-1-42* Witt, Johannes DS2-2-90* Vujasinovic, Todor A-02 Wittbrodt, Beate DS2-1-28 Vuoriluoto, Karoliina DS3-3-32 Wittbrodt, Joachim DS1-2-16, DS2-1-28 Wittig, Ulrike DS1-2-05 W Wittmann, Dominik DS2-2-11* Wiuf, Carsten DS2-2-04*, DS2-3-03 Wadhams, George A-35 Wodke, Judith DS1-1-47 Wagenmakers, Anton DS2-1-15 Woelfl, Stefan DS2-4-14 Wagner, Andreas DS2-2-104, P-24* Wolf, Denise DS2-4-24 Waldon, Sally A. DS1-2-03 Wölfl, Stefan DS1-1-63 Wallace, Douglas DS2-2-66 Wolkenhauer, Olaf A-25*, DS2-1-109, DS2-1-37, Wallman, Mikael OS-10* DS2-1-38, DS2-1-80, Wallqvist, Anders DS2-2-70 DS2-2-52, DS2-2-77, OS-43 Waltermann, Christian DS2-1-111, DS2-2-19* Wolski, Eryk DS2-2-107 Walther, Dirk DS1-4-18 Workman, Chris A-07 Wan, Frederic DS2-1-89 Wren, Brendan DS2-2-133 Wang, Horng-Dar OS-14 Wrzosek, Antoni OS-37 Wang, Lu DS2-1-86 Wu, Boqian DS2-1-104 Wang, Ruiqi DS2-2-84* Wu, Guanming DS1-2-04 Wanker, Erich A-11 Wu, Hsiao-Huei DS3-1-16 Wanner, Barry DS2-4-13*, OS-23 Wu, Jake DS2-2-107 Wardleworth, Leanne DS1-1-32 Wurst, Wolfgang DS2-2-11 Warringer, Jonas DS1-1-18, DS2-1-111, Wyller, John A. DS2-2-72 DS2-1-68*, DS2-4-45 Watahiki, Masaaki K. DS1-4-09 X Waters, N. DS1-4-21 Waters, S. DS1-4-21 Xie, Xiao Liang Sunney P-11* , DS3-1-08

Wattis, Jonathan DS1-1-08 Xing, Heming DS2-1-70 Index Author Weckström, Matti DS2-1-61 Weckwerth, Wolfram DS1-4-18 Y Wedelin, Dag DS1-2-10 Wegner, Katja A-14*, DS2-2-103* Yachie-Kinoshita, Ayako DS2-2-15 Weichart, Dieter A-04*, DS1-1-65* Yaffe, Mike DS2-1-24, P-05 Weidemann, Andreas DS1-2-05 Yamaguchi, Rui DS2-1-50 Weissman, Irving DS3-4-24 Yamamoto, Kotaro T. DS1-4-09 Weith, Andreas DS1-3-26 Yamamoto, Natsuko OS-23 Welham, Patricia DS2-2-82* Yamamoto, Tadashi DS2-1-117, DS2-1-119 Wennberg, Bernt DS3-3-43 Yamauchi, Mai DS2-1-50* West, K. A. P-19 Yan, Ching-Cher Sanders DS3-4-11 Wester, Katja DS1-4-03, DS1-4-15, Yanagida, Toshio DS2-2-36, DS3-4-23 DS1-4-23 Yanagihara, Fusamitsu OS-23 Westerhoff, Hans V. DS1-1-61, DS1-1-65, Yanai, Hagai DS2-2-34 DS2-1-73, DS2-1-81, Yang, Chin-Rang DS2-1-43* DS2-1-93, DS2-4-14*, Yang, Lei DS2-4-08* DS1-3-01*, A-04, A-06, A-27, Yang, Tae Hoon DS2-4-21 DS1-1-01, DS1-1-16, Ye, Tian DS2-1-109 DS1-1-63, DS1-3-08, Yeh, Pamela DS2-4-05 DS2-4-10, OS-02 Yeo, Zhenxuan DS3-1-13 Westermark, Pål DS2-1-36* Yetukuri, Laxman DS1-1-02, DS2-3-15* Westly, Elizabeth DS2-3-06 Yeturu, Kalidas DS1-3-25* Westoll, Julian DS1-1-19* Yevshin, Ivan DS1-1-67, DS1-2-19, Westra, Ronald DS2-2-75 DS2-1-96, DS2-2-149 Whelan, Ken DS3-4-08 Yi, Tau-Mu DS2-1-113, DS2-1-79 Whitaker, Dawn R DS2-4-13 Ying, Bei-Wen DS2-1-05, DS2-1-114 White, Michael DS1-2-15, DS2-1-02*, Yokoi, Kazuhito DS2-1-16 DS2-1-55 Yokota, Jun DS2-1-50 Wickstead, Bill DS2-4-44 Yomo, Tetsuya DS2-1-05*, DS2-1-114, Wiebe, Marilyn DS1-1-68 DS3-4-10 Wiechert, Wolfgang DS2-2-43 Yoo, Young Sook DS2-2-151 Wierling, Christoph A-31* Yoshida, Ryo DS2-1-50 Wiersma, Anne DS2-4-27 Yoshida, Shimuzu OS-11 Wilczynski, Bartek DS2-1-110 Yoshida, Yuki DS1-3-12*, DS3-4-22 Wild, David L. DS1-4-20 Yoshimasa, Tsuruoka A-51 Wilkinson, Anthony A-30 Young, Mike DS3-4-08 Wilkinson, Darren A-46, DS2-2-107* Youssef, Simon DS3-4-09* Wilkinson, Steve DS2-4-14 Yu, Mei DS2-1-07 Willmitzer, Lothar DS1-1-50 Yu, Richard DS2-4-04*, DS3-1-05

ICSB 2008 229 Yuh, Chiou-Hwa OS-14* Yumoto, Noriko DS2-1-116 Yus, Eva DS1-1-47*, DS2-1-56 Yvert, Gael P-26*

Z

Zagaris, Antonios DS2-2-148 Zakhartsev, Maksim DS2-4-14 Zamboni, Nicola DS1-1-46, DS1-1-66, DS2-3-24, P-13 Zamborszky, Judit DS2-1-08* Zampieri, Mattia DS1-1-41, DS2-2-53, DS2-3-11 Zandstra, Peter DS2-1-07 Zanghellini, Juergen DS2-4-31 Zapatka, Marc DS1-3-26 Zarzer, Clemens DS2-1-87 Zavolan, Mihaela DS2-1-124, DS2-1-76 Zeef, Leo A. DS1-1-32 Zeemering, Stef DS2-2-75* Zeevi, Yehoshua DS2-2-110 Zerial, Marino DS2-1-98 Zhang, Bing DS1-1-33 Zhang, Chaoyang DS3-4-13 Zhang, Jie DS2-1-39* Zhang, Nianshu DS1-1-32 Zhang, Qiang DS2-2-128 Zhang, Zhongge DS2-4-19, DS2-4-36 Zhen Xuan, Yeo DS2-2-87 Zheng, Yanan DS3-3-21 Zhu, Qingwei DS2-1-23

Author Zi, Zhike DS3-3-21* Index Zilles, Karl A-11 Zinovyev, Andrei DS2-1-57, DS2-2-106* Zupan, Blaz DS2-1-53

230 ICSB 2008