Regularized by Physics: Graph Neural Network Parametrized Potentials for the Description of Intermolecular Interactions

Abstract

Simulations of molecular systems using electronic structure methods are still not feasible for many systems of biological importance. As a result, empirical methods such as force fields (FF) have become an established tool for the simulation of large and complex molecular systems. The parametrization of FF is, however, time-consuming and has traditionally been based on experimental data. Recent years have therefore seen increasing efforts to automatize FF parametrization or to replace FF with machine-learning (ML) based potentials. Here, we propose an alternative strategy to parametrize FF, which makes use of ML and gradient-descent based optimization while retaining a functional form founded in physics. Using a predefined functional form is shown to enable interpretability, robustness, and efficient simulations of large systems over long time scales. To demonstrate the strength of the proposed method, a fixed-charge and a polarizable model are trained on ab initio potential-energy surfaces. Given only information about the constituting elements, the molecular topology, and reference potential energies, the models successfully learn to assign atom types and corresponding FF parameters from scratch. The resulting models and parameters are validated on a wide range of experimentally and computationally derived properties of systems including dimers, pure liquids, and molecular crystals.

Hybrid classical/machine-learning force fields for the accurate description of molecular condensed-phase systems

“…The notion of atom types used as part of the proposed model relies on the formalism proposed in ref. 35. This formalism makes use of atom types, which are learned from molecular graphs, i.e., graphs that do not include information about the geometry of a molecule but only its covalent bonds.…”

“…Here, the model is re-trained and improved based on the model architecture described in ref. 61. Implementation of the electrostatic interaction and Ewald summation follows the formalism outlined in refs.…”

“…60 was replaced with the AMP formalism described in ref. 61. For training, the dataset generated in ref.…”

“…The former takes advantage of the automatic differentiation based parametrization framework described in previous work. 35 Automatic differentiation has emerged as a powerful tool in computational science, permitting efficient gradient-based parametrization of physical models. [36][37][38] The latter is used to introduce a higher degree of exibility, which is necessary for situations where classical approximations break down, for instance at short distances and large overlaps.…”

End-to-End Differentiable Force Field Generator with Crystal Structure Differentiation and Matching

“…This is important for small organic molecules, because it contains intermolecular structure data. The reproducibility of crystal structures has been generally used to validate FFs [42][43][44]. , whereas the cohesive properties (e.g., lattice energy) have been considered to discuss the stability or strength of the intermolecular interactions [45][46][47].…”