Moonlight

Some proteins perform several apparently unrelated functions in the cell or the organism. These multifunctional proteins are collectively named 'moonlighting proteins', referring in the human society to a second job held in addition to a regular one. Because protein function is mostly investigated experimentally using hypothesis-driven approaches, the discovery of the moonlighting function(s) of a protein mainly results from the unexpected convergence of different inquiries. As a result, very few moonlighting proteins have been identified (~50 in humans), despite a growing interest for this functional singularity. For instance, an increasing number of endocytic proteins appear to regulate gene expression and several cytoplasmic proteins have been demonstrated to display another function than their regular one when exported at the cell surface of tumor cells. Dedicated discovery methods are thus needed to identify moonlighting proteins. To this end, we will propose a bioinformatic approach providing this opportunity in an unbiased manner at the scale of the whole proteome. The present project aims at identifying moonlighting proteins by investigating their interaction partners in the protein-protein interaction network. Indeed, proteins insure their function through interactions (physical contacts) with other proteins and the set of all these interactions in the cell forms a huge protein-protein interaction network. Given that moonlighting proteins should interact with different sets of proteins to perform their different functions, we need to recognize in the network proteins at the intersection of protein sets, each set formed by functionally related proteins. As explain in project, we will: - develop novel graph partitioning algorithms and methods allowing to classify vertices/proteins in overlapping classes (WP1); - mine the semantic representation of protein function (i.e. Gene Ontology) using statistical methods in order to define associated and 'repulsive' GO terms among protein annotations. This should help identify the possible cellular functions of moonlighting proteins (WP2); - apply the method (derived from WP1&2) in large-scale studies and predict the moonlighting proteins in several eukaryotic organisms from the investigation of their interaction networks (WP3). Ultimately, the results of this work, at the interface between biology and computer science, should contribute to build novel hypotheses experimentally testable by proposing proteins that are potentially multifunctional.