2026-04-17T07:32:45Zhttps://recercat.cat/oai/request

oai:recercat.cat:10230/711992025-09-17T14:31:54Zcom_2072_6col_2072_452952

00925njm 22002777a 4500 dc Charron, Nicholas E. author Pérez Culubret, Adrià author Majewski, Maciej author De Fabritiis, Gianni author Clementi, Cecilia author 2025-09-16T05:49:11Z 2025-09-16T05:49:11Z 2025 The most popular and universally predictive protein simulation models employ all-atom molecular dynamics, but they come at extreme computational cost. The development of a universal, computationally efficient coarse-grained (CG) model with similar prediction performance has been a long-standing challenge. By combining recent deep-learning methods with a large and diverse training set of all-atom protein simulations, we here develop a bottom-up CG force field with chemical transferability, which can be used for extrapolative molecular dynamics on new sequences not used during model parameterization. We demonstrate that the model successfully predicts metastable states of folded, unfolded and intermediate structures, the fluctuations of intrinsically disordered proteins and relative folding free energies of protein mutants, while being several orders of magnitude faster than an all-atom model. This showcases the feasibility of a universal and computationally efficient machine-learned CG model for proteins. We thank all members of the Clementi and Noé groups for their help in different phases of this work. We gratefully acknowledge funding from the European Commission (grant no. ERC CoG 772230 ‘ScaleCell’), the International Max Planck Research School for Biology and Computation (IMPRS–BAC), the Bundesministerium für Bildung und Forschung BMBF (Berlin Institute for Learning and Data, BIFOLD, and project FAIME 01IS24076), the Berlin Mathematics Center MATH+ (AA1-6, EF1-2) and the Deutsche Forschungsgemeinschaft DFG (NO 825/2, NO 825/3, NO 825/4; SFB/TRR 186, Project A12; SFB 1114, Projects B03, B08 and A04; SFB 1078, Project C7; and RTG 2433, Projects Q05 and Q04), the National Science Foundation (CHE-1900374 and PHY-2019745) and the Einstein Foundation Berlin (Project 0420815101). We gratefully acknowledge the computing time provided on the supercomputer Lise at NHR@ZIB as part of the NHR infrastructure, and on the supercomputer JUWELS operated by the Jülich Supercomputing Centre. We thank volunteers at GPUGRID.net for contributing computational resources and Acellera for funding. http://hdl.handle.net/10230/71199 Computational biophysics Computational chemistry Navigating protein landscapes with a machine-learned transferable coarse-grained model