Navigating protein landscapes with a machine-learned transferable coarse-grained model

• DOI: 10.48550/arxiv.2310.18278
• Publication Date: 2023-01-01
• Authors: Nicholas E. Charron; Félix Musil; Andrea Guljas; Yaoyi Chen; Klara Bonneau; Aldo S. Pasos-Trejo; Jacopo Venturin; Daria Gusew; Iryna Zaporozhets; Andreas Krämer; Clark Templeton; Atharva Kelkar; Aleksander E. P. ...; Simon Olsson; Adrià Pérez; Maciej Majewski; Brooke E. Husic; Ankit Patel; Gianni De Fabritiis; Frank Noé; Cecilia Clementi
• Keywords: Parametrization (atmospheric modeling); Transferability; Metastability; Force Field

The most popular and universally predictive protein simulation models employ all-atom molecular dynamics (MD), but they come at extreme computational cost. development of a universal, computationally efficient coarse-grained (CG) model with similar prediction performance has been long-standing challenge. By combining recent deep learning methods large diverse training set simulations, we here develop bottom-up CG force field chemical transferability, which can be used for extrapolative on new sequences not during parametrization. We demonstrate that the successfully predicts folded structures, intermediates, metastable unfolded basins, fluctuations intrinsically disordered proteins while it is several orders magnitude faster than an model. This showcases feasibility universal machine-learned proteins.