Interpreting Molecular Dynamics Forces as DeepLearning Gradients Improves Quality Of PredictedProtein Structures

Abstract

Protein structure predictions from deep learning models like AlphaFold2, despite their remarkable accuracy, are likely insufficient for direct use in downstream tasks like molecular docking. The functionality of such models could be improved with a combination of increased accuracy and physical intuition. We propose a new method to train deep learning protein structure prediction models using molecular dynamics force fields to work toward these goals. Our custom PyTorch loss function, OpenMM-Loss, represents the potential energy of a predicted structure. OpenMM-Loss can be applied to any all-atom representation of protein structure capable of mapping into our software package, SidechainNet. We demonstrate our methods efficacy by finetuning OpenFold. We show that subsequently predicted protein structures, both before and after a relaxation procedure, exhibit comparable accuracy while displaying lower potential energy and improved structural quality as assessed by MolProbity metrics. SIGNIFICANCEWe propose a novel framework to directly incorporate forces from molecular dynamics as gradients for training deep learning models like AlphaFold2. We implement our method as a PyTorch loss function, OpenMM-Loss, which frames the potential energy of predicted protein structures as a minimization objective. When applied to OpenFold, our method demonstrates improved structural quality, lower potential energy, and comparable accuracy relative to the OpenFold baseline. Our framework may enhance the ability of deep learning models to recapitulate fundamental biophysical principles, reducing the number of structural irregularities in their predictions and paving the way for more effective downstream applications.