Protein Diversity Generation: from Evolution to Design – ProDiGen

The need to evolve in response to environmental changes has led living organisms to acquire powerful diversification strategies. Among these, Diversity Generating Retro-elements (DGRs) are natural directed mutagenesis systems capable of efficiently exploring sequence space. DGRs have been identified in bacteriophages that use this diversification system to evolve their attachment protein to the bacterial receptor, allowing them to change host. In this project, we propose to first study DGR systems in order to quantitatively characterize the extent to which they can generate a wide variety of proteins with appropriate functional and structural properties. This objective will require establishing new models to score protein sequences using unsupervised machine learning techniques that we will train on available sequence databases. Then, we will develop a mutagenesis system based on DGRs, working synergistically with the computational models through an iterative process of active learning, to design new proteins in an optimized manner. We will test this procedure on the Cas9 protein, a nuclease with numerous applications as a research tool as well as in medicine, in order to design variants capable of recognizing PAM nucleotide motifs different from the natural ones. Our project therefore has the dual objective of advancing the understanding of a fundamental biological mechanism and exploiting it to design proteins with novel properties, a field with important applications in industry. It is based on a close collaboration between a computational team at the ENS and an experimental team at the Institut Pasteur.