Two for One: Diffusion Models and Force Fields for Coarse-Grained Molecular Dynamics

• DOI: 10.1021/acs.jctc.3c00702
• Publication Date: 2023-09-09T13:08:15Z
• Authors: Marloes Arts; Victor Garcia Sat...; Chin-Wei Huang; Daniel Zügner; Marco Federici; Cecilia Clementi; Frank Noé; Robert Pinsler; Rianne van den Berg
• Keywords: Force Field; Leverage (statistics); Dynamics

Coarse-grained (CG) molecular dynamics enables the study of biological processes at temporal and spatial scales that would be intractable an atomistic resolution. However, accurately learning a CG force field remains challenge. In this work, we leverage connections between score-based generative models, fields, to learn without requiring any inputs during training. Specifically, train diffusion model on protein structures from simulations, show its score function approximates can directly used simulate dynamics. While having vastly simplified training setup compared previous demonstrate our approach leads improved performance across several simulations for systems up 56 amino acids, reproducing equilibrium distribution preserving all-atom such as folding events.