Trajectory-based training enables protein simulations with accurate folding and Boltzmann ensembles in cpu-hours

An ongoing challenge in protein chemistry is to identify the underlying interaction energies that capture protein dynamics. The traditional trade-off in biomolecular simulation between accuracy and computational efficiency is predicated on the assumption that detailed force fields are typically well-parameterized, obtaining a significant fraction of possible accuracy. We re-examine this trade-off in the more realistic regime in which parameterization is a greater source of error than the level of detail in the force field. To address parameterization of coarse-grained force fields, we use the contrastive divergence technique from machine learning to train from simulations of 450 proteins. In our procedure, the computational efficiency of the model enables high accuracy through the precise tuning of the Boltzmann ensemble. This method is applied to our recently developed Upside model, where the free energy for side chains is rapidly calculated at every time-step, allowing for a smooth energy landscape without steric rattling of the side chains. After this contrastive divergence training, the model is able to de novo fold proteins up to 100 residues on a single core in days. This improved Upside model provides a starting point both for investigation of folding dynamics and as an inexpensive Bayesian prior for protein physics that can be integrated with additional experimental or bioinformatic data.Author Summary: All-atom biomolecular simulations are useful but often take months to simulate biologically relevant reactions. Coarse-grain folding simulations reduce the computational requirements; however, they typically have reduced accuracy and knowledge of the native state is required. Here, we show that a properly formulated coarse-grain model trained using modern machine learning methods can rival all-atom models for de novo protein folding and dynamics simulations. The Upside model’s success argues that simpler models that can be globally parameterized can rival more detailed but slower models whose parameterization is more challenging—more complexity does not necessarily equate to higher accuracy. Upside’s ready generation of Boltzmann ensembles allows for a wide range of computational studies of protein folding, dynamics and binding. Additionally, in studies that incorporate experimental or bioinformatics data, including sparse contact predictions, Upside provides an inexpensive Bayesian prior distribution over protein structures that may be updated using experimental information.