D'une représentation statique à un modèle dynamique des réseaux d'interaction protéine-protéine

To further understand biological processes, it is important to consider gene functions in the context of complex molecular networks. I joined the lab of Marc Vidal in January 2001, in the midst of its effort to decipher at the scale of the proteome the complex network of protein-protein interactions in the metazoan model organism C. Elegans. To accomplish such a task, one has to first generate a physical resource of coding sequences that can be used in yeast two-hybrid (Y2H) screenings, as well as various other functional assays. In this respect, I participated in the development of a bioinformatics platform facilitating the cloning of 12,000 of the 19,000 predicted Open Reading Frames (ORFs) in C. Elegans (Nature Genetics 2003). This platform was subsequently used in two similar efforts to generate the ORFeome of the pathogen bacteria B. Melitensis (Genome Research 2004a) (96% of the predicted ORFs of B. Melitensis were cloned) and a first version of the human ORFeome containing approximately 8,000 ORFs (Genome Research 2004b). Using the C. Elegans ORFeome resource, a huge team effort led to the generation of one of the first metazoan interactomes, uncovering the network formed by 5,500 protein-protein interactions (Science 2004). My involvement in that project was the development of a bioinformatics pipeline allowing an efficient acquisition of the data generated by the high-throughput Y2H screenings, as well as tools to integrate this dataset with the huge body of transcriptional and phenotypic data available. I have demonstrated that currently known protein-protein interactions cover only a small portion of the full interactomes (Nature Biotechnology 2005). Nevertheless, they can be used as a scaffold on which other functional information can be overlaid to improve our understanding of key biological processes. By creating tools to bring together genetic, transcriptional and functional data into the interactome, I participated in the discovery of new functional links between genes expressed in the germline of C. Elegans (Current Biology 2002), as well as the uncovering of new components involved in the TGF-beta signaling pathway (Molecular Cell 2004), the RNAi machinery (Science 2005) and C. Elegans embryogenesis (Nature 2005). All of these studies generated testable hypotheses that strengthened the implication of those new functional links in each of the biological processes investigated. The integration of functional datasets can also be used to reveal emergent properties of biological networks. Taking advantage of the wealth of proteomic, transcriptional and genetic data available for S. Cerevisiae, I have shown that the genetic robustness of the yeast is linked to the modular and hierarchical nature of the topology of its interactome (Nature 2004 and recent submission PloS Biology). While the static nature of current interactomes can be partially overcome by integrating transcriptome and interactome data in unicellular organisms (Nature 2004 and recent submission to PloS Biology), such approaches remain limited for multicellular organisms such as C. Elegans. To decipher the dynamics of the worm interactome, one has to determine the localizations of the expression of its genes in space and in time. Therefore, I joined a high-throughput genome wide expression localization project. The originality of this project resides in a standardized and high-throughput data acquisition of in vivo gene reporter assays allowing the gathering of both spatial as well as temporal expression patterns for nearly 2,000 C. Elegans promoters. This new biological map, the "localizome", will be used to not only refine the current static interactome and define tissue-specific interactomes, but also to gain a system-level understanding of gene regulation during the post-embryonic development of C. Elegans (under consideration in Nature)