TY - JOUR A1 - Charron, Nicholas A1 - Musil, Félix A1 - Guljas, Andrea A1 - Chen, Yaoyi A1 - Bonneau, Klara A1 - Pasos-Trejo, Aldo A1 - Jacopo, Venturin A1 - Daria, Gusew A1 - Zaporozhets, Iryna A1 - Krämer, Andreas A1 - Templeton, Clark A1 - Atharva, Kelkar A1 - Durumeric, Aleksander A1 - Olsson, Simon A1 - Pérez, Adrià A1 - Majewski, Maciej A1 - Husic, Brooke A1 - Patel, Ankit A1 - De Fabritiis, Gianni A1 - Noé, Frank A1 - Clementi, Cecilia T1 - Navigating protein landscapes with a machine-learned transferable coarse-grained model JF - Arxiv N2 - The most popular and universally predictive protein simulation models employ all-atom molecular dynamics (MD), 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 parametrization. We demonstrate that the model successfully predicts folded structures, intermediates, metastable folded and unfolded basins, and the fluctuations of intrinsically disordered proteins while it is 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. Y1 - 2023 U6 - https://doi.org/https://doi.org/10.48550/arXiv.2310.18278 ER -