Théorique, numérique
Understanding the connection between a protein's amino acid sequence and its function remains a significant challenge. Traditional biophysics approach address the 'sequence→3D structure' problem but not the 'sequence→function' problem. Recent data-driven approaches analyze protein evolution to build statistical models of this relationship. These models, rooted in statistical physics and machine learning, enable the design of new functional protein sequences. A current challenge is imbuing physical interpretability into these models, to understand how sequences correspond to different physical properties of proteins, and to design proteins with specific properties. The goal of the internship is to advance these models in this direction.