Learning neural network potentials from experimental data via Differentiable Trajectory Reweighting

Abstract

In molecular dynamics (MD), neural network (NN) potentials trained bottom-up on quantum mechanical data have seen tremendous success recently. Top-down approaches that learn NN potentials directly from experimental data have received less attention, typically facing numerical and computational challenges when backpropagating through MD simulations. We present the Differentiable Trajectory Reweighting (DiffTRe) method, which bypasses differentiation through the MD simulation for time-independent observables. Leveraging thermodynamic perturbation theory, we avoid exploding gradients and achieve around 2 orders of magnitude speed-up in gradient computation for top-down learning. We show effectiveness of DiffTRe in learning NN potentials for an atomistic model of diamond and a coarse-grained model of water based on diverse experimental observables including thermodynamic, structural and mechanical properties. Importantly, DiffTRe also generalizes bottom-up structural coarse-graining methods such as iterative Boltzmann inversion to arbitrary potentials. The presented method constitutes an important milestone towards enriching NN potentials with experimental data, particularly when accurate bottom-up data is unavailable.

This is a post-peer-review, pre-copyedit version of an article published in Nature Communications. The final authenticated version is available online at: http://dx.doi.org/10.1038/s41467-021-27241-4

References in corpus (8)

• Machine learning for molecular simulation
• A Fourth-Generation High-Dimensional Neural Network Potential with Accurate Electrostatics Including Non-local Charge Transfer
• The Individual and Collective Effects of Exact Exchange and Dispersion Interactions on the Ab Initio Structure of Liquid Water
• TorchMD: A deep learning framework for molecular simulations
• A Differentiable Programming System to Bridge Machine Learning and Scientific Computing
• Differentiable sampling of molecular geometries with uncertainty-based adversarial attacks
• Towards empirical force fields that match experimental observables
• Learning to Control PDEs with Differentiable Physics

Cited by in corpus (3)

• Machine Learning Coarse-Grained Potentials of Protein Thermodynamics
• Deep Coarse-grained Potentials via Relative Entropy Minimization
• Learning Pair Potentials using Differentiable Simulations