Dresden 2026 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

BP: Fachverband Biologische Physik

BP 15: Computational Biophysics III

BP 15.7: Talk

Wednesday, March 11, 2026, 11:30–11:45, BAR/SCHÖ

Modelling Protein Dynamics with Machine Learned Coarse-Grained Models — •Klara Bonneau — Freie Universität Berlin, Germany

The most popular and universally predictive protein simulation models employ all-atom molecular dynamics (MD). However, understanding dynamical processes like protein folding, interactions, and aggregation requires accessing timescales beyond conventional MD capabilities. Coarse-graining (CG) accelerates simulations by focusing on essential degrees of freedom, but a universally predictive CG model for proteins remains elusive. Our work introduces the first thermodynamically consistent CG model that extrapolates to unseen protein sequences. By leveraging state-of-the-art machine learning techniques, we simulate the folding of unknown proteins, protein-protein interactions, intrinsically disordered proteins, and mutation effects. Current extensions include larger proteins and protein-ion/small molecule interactions, surpassing conventional MD timescale limitations. Additionally, explainable AI techniques enable us to interpret results and demonstrate that deep learning models capture physically consistent interactions.

Keywords: Protein Dynamics; Molecular Dynamics; Coarse-Graining; Machine-Learning; Explainable AI