Molecular Force Fields with Gradient-Domain Machine Learning (GDML): Comparison and Synergies with Classical Force Fields

TL;DR

This paper compares machine learning-based force fields with classical force fields, exploring their differences, complementarities, and ways to enhance classical force fields using insights from ML approaches.

Contribution

It demonstrates how ML-FFs can complement classical force fields and proposes reparametrizing classical force fields for improved accuracy without losing their transferability.

Findings

ML-FFs accurately predict energies and forces at high-level ab initio accuracy.

Classical force fields can be improved by reparametrization based on ML insights.

The combined approach enhances the understanding and accuracy of molecular simulations.

Abstract

Modern machine learning force fields (ML-FF) are able to yield energy and force predictions at the accuracy of high-level $ab~initio$ methods, but at a much lower computational cost. On the other hand, classical molecular mechanics force fields (MM-FF) employ fixed functional forms and tend to be less accurate, but considerably faster and transferable between molecules of the same class. In this work, we investigate how both approaches can complement each other. We contrast the ability of ML-FF for reconstructing dynamic and thermodynamic observables to MM-FFs in order to gain a qualitative understanding of the differences between the two approaches. This analysis enables us to modify the generalized AMBER force field (GAFF) by reparametrizing short-range and bonded interactions with more expressive terms to make them more accurate, without sacrificing the key properties that make MM-FFs so successful.