Chemoinformatics and bioinformatics methods are now necessary in every drug discovery program. Pharmaceutical industries dedicate more than 10% of their research and development investment in computer aided drug design (Kapetanovic 2008). The emergence of these tools can be explained by the increasing availability of high performance calculating machines and also by the low cost of in silico analysis compared to in vitro tests.Biological tests that were performed over last decades are now a valuable source of information and a lot of databases are trying to list them. This huge amount of information led to the birth of a new research field called “chemogenomics”. The latter is focusing on the identification of all possible associations between all possible molecules and all possible targets. Thus, using chemogenomics approaches, one can obtain a biological profile of a molecule and even anticipate possible side effects.This thesis was focused on the development of approaches that aim to predict the binding of molecules to targets. In our lab, we focus on profiling molecular databases in order to get their full biological profile. Thus, my main work was related to this context and I tried to develop predictive models to assess the binding of ligands to proteins, to validate some virtual screening methods for profiling purpose, and finally, I developed an automatic hybrid profiling workflow that selects the best fitted virtual screening approach to use according the ligand/target context.
