Hot Topics in Structural Genomics†
Total Page:16
File Type:pdf, Size:1020Kb
Comparative and Functional Genomics Comp Funct Genom 2003; 4: 394–396. Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/cfg.314 Conference Report The 4th Bologna Winter School: hot topics in structural genomics† Rita Casadio* Department of Biology/CIRB, University of Bologna, Via Irnerio 42, 40126 Bologna, Italy *Correspondence to: Abstract Rita Casadio, Department of Biology/CIRB, University of The 4th Bologna Winter School on Biotechnologies was held on 9–15 February Bologna, Via Irnerio 42, 40126 2003 at the University of Bologna, Italy, with the specific aim of discussing recent Bologna, Italy. developments in bioinformatics. The school provided an opportunity for students E-mail: [email protected] and scientists to debate current problems in computational biology and possible solutions. The course, co-supported (as last year) by the European Science Foundation program on Functional Genomics, focused mainly on hot topics in structural genomics, including recent CASP and CAPRI results, recent and promising genome- Received: 3 June 2003 wide predictions, protein–protein and protein–DNA interaction predictions and Revised: 5 June 2003 genome functional annotation. The topics were organized into four main sections Accepted: 5 June 2003 (http://www.biocomp.unibo.it). Published in 2003 by John Wiley & Sons, Ltd. Predictive methods in structural Predictive methods in functional genomics genomics • Contemporary challenges in structure prediction • Prediction of protein function (Arthur Lesk, and the CASP5 experiment (John Moult, Uni- University of Cambridge, Cambridge, UK). versity of Maryland, Rockville, MD, USA). • Microarray data analysis and mining (Raf- • Contemporary challenges in structure prediction faele Calogero, University of Torino, Torino, (Anna Tramontano, University ‘La Sapienza’, Italy). Rome, Italy). • Prediction of protein function and protein net- • Prediction of protein structure and function at the works (Søren Brunak Technical University of genomic scale (Jeffrey Skolnick, Buffalo Center Denmark, Lyngby, Denmark, and Alfonso Valen- of Excellence in Bioinformatics, Buffalo, NY, cia, National Centre of Biotechnology, Canto- USA). blanco, Spain). • Advanced automated machine learning appr- oaches (David Jones, University College, Lon- don, UK). • Fully automated ab initio protein structure pre- Prediction of membrane protein diction (Chris Bystroff, Rensselaer Polytechnic structure Institute, Troy, NY, USA). • Automatic fold recognition prediction (Daniel • Fischer, Ben Gurion University, Be’er Sheva, The prediction of membrane protein topology Israel). (Gunnar von Heijne, University of Stockholm, Stockholm, Sweden, and Stephen White, Uni- versity of California, Irvine, CA, USA). † This article is an adaptation of the Conference Report on this • workshop (published to satisfy the requirements of the ESF Application of structural genomics tools to fish- funding) previously published on the Integrated Approaches for ing for new membrane proteins (Rita Casadio, Functional Genomics website. University of Bologna, Italy). Published in 2003 by John Wiley & Sons, Ltd. 4th Bologna Winter School: hot topics in structural genomics 395 Prediction of protein–protein What about predicting function? and protein–DNA interaction • The CAPRI experiment and the prediction of Raffaele Calogero discussed DNA microarray tech- protein–protein interactions (Joel Janin, LEBS, nology, a high-throughput method for gaining CNRS, Gif sur Yvette, France). information on gene function. However, genomes • Prediction of protein complexes based on evolu- can only be completely annotated when we can tionary information (Patrick Aloy, EMBL Hei- predict the function, possibly starting from the delberg, Germany). sequence. This is presently a really difficult and • Prediction of protein–DNA complexes (Sue challenging task, as Arthur Lesk pointed out in Jones, European Bioinformatics Institute, Hinx- his talk. Søren Brunak has used a suite of pro- ton, UK). grams, integrating predictions of different structural • Prediction of functional patches in proteins and functional properties, to successfully address (Manuela Helmer Citterich University of Tor this problem. Alfonso Valencia described how pre- Vergata, Rome, Italy, and Nir Ben Tal, Tel Aviv dictions are of a quality similar to the exper- University, Tel Aviv, Israel). imental data, when different interacting protein networks (at the basis of systems biology, and obtained using both theoretical and experimental What can we predict in the post genomic approaches) are compared. Also, the interactions era? Is it feasible to perform large-scale detected by more than one method have a substan- prediction on entire genomes? How can tially higher confidence. we cope with predictive tools? The CASP (critical assessment of protein structure prediction techniques) experiment reports on how Good results are being achieved in the different methods perform in specific predictive case of globular proteins, but when it tasks. The results had been debated just 2 months comes to membrane proteins, the earlier at the CASP5 meeting in Asilomar on 1–5 situation is much worse December 2002 (http://predictioncenter.llnl.gov/ casp5/Casp5.html). Apparently, as explained by its organizer (John Moult) and one of its asses- Less than 1% of PDB structures are membrane sors (Anna Tramontano, assessor of the homology proteins and, as a result, building by homol- building section), one cannot state that a partic- ogy is hampered by the paucity of examples, ular method is the best. There are several excel- except possibly in the case of outer membrane lent tools that, when integrated into meta-servers proteins. The β-barrel architecture is rather con- (as described by Daniel Fisher), can perform fold served and the changes that are seen are in recognition in a very satisfactory way. Therefore, the number of antiparallel beta strands. Once integrated knowledge will help us in solving the this is predicted, a 3D model can be computed folding problem also on a genome-wide scale. Jef- with a threading procedure (as explained in my frey Skolnick is modelling most of the protein own presentation). In the case of inner mem- content of presently available known proteomes, brane proteins, one is mainly left with the pre- including some functional properties, based on a diction of topological models. Gunnar von Hei- sequence-to-structure-to-function paradigm. With a jne presented some recent advances in the iden- very low rate of false positives, his tools can also tification of membrane protein topology and its predict folds of some 30–50% of the proteins. prediction with bioinformatic methods. Stephen Machine learning approaches form the basis of the White described two Web-based tools, MPEx most successful tools for prediction of structural (http://blanco.biomol.uci.edu/mpex) and MPtopo features, including secondary structure of proteins, (http://blanco.biomol.uci.edu/mptopo), which are as David Jones reported. Chris Bystroff explained based on chemical-physical properties of trans- that proteins can also be predicted ab initio,pro- membrane α-helices and ‘designed to help experi- vided that an HMM-based method, predicting pro- mentalists explore the topology of membrane pro- tein contact maps, is used. teins of unknown 3D structure’. Published in 2003 by John Wiley & Sons, Ltd. Comp Funct Genom 2003; 4: 394–396. 396 R. Casadio Can we predict protein–protein characterization (Alessandro Desideri, University and protein–DNA interactions? of Tor Vergata, Rome, Italy, and Richard Wag- ner, University of Osnabruck,¨ Germany), protein Joel Janin described the CAPRI (critical assessment structure determination with NMR (Henriette Moli- of predicted interactions; http://capri.ebi.ac.uk) nari, University of Verona, Italy), and protein fold- experiment. Inspired by CASP, it similarly accepts ing studies with atomic force microscopy (Bruno predictions of proteins interacting in complexes, Samori, University of Bologna, Italy). whose structures are known only after the submis- The take home message was that bioinformatics sion is closed. The results of the last edition (which can offer solutions to problems, provided that were evaluated by Shoshana Wodak and Raul theoreticians and experimentalists work in close Mendez, Free University of Brussels, Belgium) collaboration to benefit from both a computational show significant success on some of the targets. and an experimental approach. However, the predictions failed on other targets, and Joel believes that progress is necessary ‘in the Acknowledgements score functions, and in the way docking procedures handle conformation changes and non-structural The school was made possible thanks to the invaluable information, before large-scale predictions of pro- help of Piero Fariselli, Pier Luigi Martelli, Ivan Rossi, Gianluca Tasco, Emidio Capriotti, Luca Malaguti, Remo tein–protein interactions can be made reliably’ Calabrese and Mario Compiani, all at the Biocomputing (http://www.biocomp.unibo.it/school/html2003). Unit, and the precious collaboration of Professor Lanfranco Sue Jones detailed some interesting features of pro- Masotti (Co-chairman of the School) at the Department tein–DNA complexes and Patrick Aloy explained of Biochemistry, University of Bologna. Other supporting how protein interactions can be predicted through organizations were: the University of Bologna, the Interde- tertiary structure. If we know the structure, we can partmental Centre for Biotechnological