Differentiable Rotamer Sampling with Molecular Force Fields

TL;DR

This paper develops a mathematical foundation for Boltzmann generators, enabling efficient neural network-based sampling of molecular conformations, potentially replacing traditional molecular dynamics in complex biological systems.

Contribution

It provides a rigorous theoretical framework and practical toolkit to enhance Boltzmann generators, making them more feasible for complex macromolecular simulations.

Findings

Boltzmann generators can be faster than traditional MD for certain proteins.

The paper establishes a mathematical basis for neural network sampling in molecular systems.

A comprehensive toolkit for energy landscape exploration is introduced.

Abstract

Molecular dynamics is the primary computational method by which modern structural biology explores macromolecule structure and function. Boltzmann generators have been proposed as an alternative to molecular dynamics, by replacing the integration of molecular systems over time with the training of generative neural networks. This neural network approach to MD samples rare events at a higher rate than traditional MD, however critical gaps in the theory and computational feasibility of Boltzmann generators significantly reduce their usability. Here, we develop a mathematical foundation to overcome these barriers; we demonstrate that the Boltzmann generator approach is sufficiently rapid to replace traditional MD for complex macromolecules, such as proteins in specific applications, and we provide a comprehensive toolkit for the exploration of molecular energy landscapes with neural networks.