4 Chemogenomics Chapter 1 Chapter 2 Chapter 3 Chapter 4 Chapter 5 Chapter 6 Quorum Chemo- Descriptor Spial sensing genomics relationships Conclusions Introduction and perspectives Atomic Pathway Proteome Cellular level level level level Chapter 4: Chemogenomics 4-1 Contents of chapter 4 Contents of chapter 4.................................................................................................. 1 Summary of chapter 4................................................................................................. 2 Introduction ................................................................................................................. 2 Chemogenomic experiments................................................................................... 4 Heterozygous deletions ....................................................................................... 5 Homozygous deletions......................................................................................... 5 Overexpression.................................................................................................... 6 Beyond deletion and overexpression libraries ..................................................... 7 Detection methods................................................................................................... 7 Non-competitive arrays........................................................................................ 7 Competitive mutant pools .................................................................................... 7 Data interpretation and analysis.............................................................................. 8 Clustering............................................................................................................. 8 Matrix operations ................................................................................................. 9 Applications ........................................................................................................... 10 Chemogenomic networks.......................................................................................... 12 Results and discussion.......................................................................................... 13 Interpretation of a chemogenomic network........................................................ 14 Properties of the chemogenomic network.......................................................... 15 Materials and methods .......................................................................................... 16 Data acquisition ................................................................................................. 16 Construction of the chemogenomic network...................................................... 16 Computation of network parameters and overlap with other networks .............. 17 Linking the chemical and biological properties of small molecules ........................... 17 Results and discussion.......................................................................................... 18 What differentiates bioactive small molecules from others? .............................. 18 What differentiates small molecules that hit S. cerevisiae from those that hit Sc. pombe?.............................................................................................................. 20 Conclusions ....................................................................................................... 22 Materials and Methods .......................................................................................... 22 References................................................................................................................ 23 Chapter 4: Chemogenomics 4-2 Summary of chapter 4 A robust knowledge of the interactions between small molecules and specific proteins aids the development of new biotechnological tools, the identification of drug targets, and the gain of specific biological insights. Such knowledge can be obtained through chemogenomic screens. In these screens, each small molecule from a chemical library is applied to each cell type from a library of cells, and the resulting phenotypes are recorded. Chemogenomic screens have recently become very common and will continue to generate large amounts of data. Therefore methods to investigate these data become important. In this chapter, I develop computational methods for the analysis of chemogenomic data. For this, I focus on the fungus Saccharomyces cerevisiae, and, to a lesser extent, Schizosaccharomyces pombe. Parts of this chapter appear in the following articles: Wuster, A. and M. Madan Babu (2008). “Chemogenomics and biotechnology.” Trends Biotechnol 26(5): 252-8. Wuster, A. (2008). “Networking with drugs.” Mol BioSyst 4(1): 14-7 The section on linking the chemical and biological properties of SMs would not have been possible without the help of Laura Schuresko-Kapitzky and Nevan Krogan of the University of California, San Francisco, who generated much of the data this section is based on. Introduction The effects of small molecules (SMs) on cells were central to the research of Paul Ehrlich (1854-1915; figure 4-1). For much of his career he strived to identify 'magic bullets', SMs that would allow him to target specific tissues or microbes whilst sparing others (Ehrlich 1911). In order to find a cure for syphilis, he systematically screened a library of hundreds of SMs for their effect against Treponema pallidum, the causative agent of the disease. After testing 605 different compounds, he eventually identified arsphenamine, which he later marketed as Salvarsan 606 (Gensini, Conti et al. 2007). With this strategy, he initiated a whole new approach to drug discovery which has persisted until today (Lipinski and Hopkins 2004). The aim of Paul Ehrlich's chemical screen was to find an SM that was active against a single pathogen. The data Chapter 4: Chemogenomics 4-3 structure created by his screen was one-dimensional and can be encoded by a single vector of the 606 compounds tested. Nowadays, chemical screens are more complex and wider in scope. In this chapter, I mostly work with the data obtained through two- dimensional chemogenomic screens. As in Paul Ehrlich's screen, the first dimension in such screens is a chemical library. The second dimension is a library of different cell types. For the data I work with in this chapter, the cell types are well-defined mutants in a library of S. cerevisiae deletion strains, where in each strain a different gene has been deleted. Alternatively, the cell types could be defined in other ways, such as in a library of cancer cell lines or of meiotic recombinants (Perlstein, Deeds et al. 2007; Perlstein, Ruderfer et al. 2007). The resulting data structure is a two- dimensional matrix where each data point has two coordinates and one specific associated value. The chemical coordinate specifies the SM that was applied, whilst the genetic coordinate specifies the cell type. The value of each data point is a measurement of the phenotype of interest, such as viability, growth rate, or cell size and shape. The results of a large number of chemogenomic screens with different designs have been published (table 4-B). They all have in common the data structure described above, but the experimental designs and aims, such as the identification of cellular targets of SMs, or the characterisation of cellular pathways, vary between them. In this chapter, I describe the different ways in which chemogenomic data can be created, and then introduce the methods that I have developed to extract information from this data. For this purpose, I have divided the results of this chapter into two parts. The first part, on chemogenomic networks, takes a gene-centric view, whilst the second part, on linking the chemical and biological properties of SMs, takes an SM-centric view. I have only tested the methods in this chapter on S. cerevisiae, and, to a lesser extent, on Sc. pombe. Nevertheless, they can potentially also be applied to other systems, such as humans. Figure 4-1. Paul Ehrlich (1854-1915) on a former D-Mark banknote. The molecular structure to the left of the portrait symbolises arsphenamine. The six-membered ring in the centre consists of arsenic atoms. It was not shown until 16 years after the issue of the banknote that arsphenamine is actually a mixture of compounds with three- and five-membered arsenic rings (Lloyd, Morgan et al. 2005). Source of the image: http://tinyurl.com/l3r9np Chapter 4: Chemogenomics 4-4 Chemogenomic experiments An in-depth understanding of the subtly different ways in which chemogenomic experiments are designed is necessary in order to develop the appropriate methods to extract meaningful information from them. Although for all methods of carrying out chemogenomic screens the resulting data structure is similar, its interpretation depends on the design of the experiment. For yeast, there are at least three different types of mutant libraries that can be generated, such as heterozygous deletions, homozygous deletions, and overexpression libraries (table 4-1). Screen A. Homozygous deletants B. Heterozygous C. Overexpression design → deletants ↓ Detection Non- competitive (Baetz, McHardy et al. 2004) arrays 1 SM > 5000 mutants aim: determine mode of action (Luesch, Wu et al. 2005) (Dudley, Janse et al. 2005) for drugs 1’000 SMs 21 conditions 7’296 mutants 4’710 mutants (Chang, Bellaoui et al. 2002) aim: identify protein targets for aim: determine gene function