Résumé : Phages are the most abundant biological entities on Earth. They modulate bacterial populations either by lysing their hosts or by conferring selective advantages by contributing phage-encoded fitness factors. Consequently, phages play a critical role in host survival and pathogenicity, and in nutrient redistribution. The analysis of phage genomes in any environment provides insights into the physiological impact of viruses on the microbial community and human health. However, in most environments the function of over 70%, and sometimes greater than 95%, of phage-encoded genes cannot be predicted based on similarity to genes with known function currently in databases. This may be due either to extreme divergence from a common ancestor or to the presence of previously unidentified functions, structures, and/or protein folds. Accurate identification of viral-encoded functions would greatly improve annotation and understanding the contribution of phages in any environment.
We have developed machine learning tools that predict gene function independently of amino acid sequence alignments. The resulting predictions are being validated using structural methods (X-ray crystallography and EM), which have unveiled new protein folds. We are combining these approaches with phenotypic responses, biochemistry, and metabolomics to probe the functions of non-structural proteins encoded by phages, and uncovering new contributions to bacterial lifestyles, with implications for microbial communities and human health.
Lieu : Auditorium - Bâtiment 21, Campus de gif