Frederic Cazals
Frederic.Cazals@inria.fr
Centre Inria d’Université Côte d’Azur, Algorithms-Biology-Structure
Abstract
The function of proteins relies on a subtle mix between structural and dynamical properties (thermodynamics, kinetics). Deep learning based methods, pioneered by AlphaFold, for which the 2024 Nobel prize in chemistry was co-awarded, have revolutionized protein structure prediction and enabled the prediction of whole proteomes.
This talk will put current work in perspective in two respects. First, I will present recent work on the critical assessment of AlphaFold predictions, stressing the gap remaining between structure and dynamics. Second, I will describe structure generation methods combining advanced kinematics and sampling techniques, as well as novel models for joint distributions of torsion angles, to make a stride towards more efficient thermodynamics and kinetics.
A quick software demo using plugins from the Structural Bioinformatics Library will conclude the talk.