Coarse-graining is an appealing strategy for simulating complex biological systems by reducing the system size and computational cost. We recently introduced a new variational method called potential contrasting for learning coarse-grained force fields and applied it for multi-scale coarse-graining of protein folding [1]. Potential contrasting generalizes the noise contrastive estimation method to formulate force field parameterization into a classification problem. Input for the method is a target ensemble of protein conformations from all-atom simulations, and no atomic force information is required. When applied to the peptide Trp-cage, we found that potential contrasting can produce force fields that accurately reproduce the all-atom conformational ensemble and capture the complex folding landscape. The method also revealed the importance of including many-body potentials in CG models to describe protein biophysics with a reduced degree of freedom and implicit solvation. In addition, we showed that potential contrasting is computationally efficient and trivially parallelizable, enabling the parameterization of transferable force fields using large datasets collected from multiple proteins.

Contrastive learning of coarse-grained force fields

Where thousands of chemistry professionals meet to share ideas and advance scientific and technical knowledge. Spring 2023 abstracts, posters, and presentations will launch at the end of March.

ACS Spring 2023

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.