Committor guided estimates of molecular transition rates

TL;DR

This paper demonstrates that neural network-based committor functions, combined with improved sampling and objective strategies, can accurately predict molecular transition rates, advancing computational methods for high-dimensional reaction analysis.

Contribution

It introduces a refined approach to learning the committor function using neural networks, enhancing accuracy and efficiency for predicting molecular transition rates.

Findings

Neural network committors accurately predict transition pathways.

Adaptive sampling improves the quality of the committor representation.

Direct application of the Hill relation yields precise transition rate estimates.

Abstract

The probability that a configuration of a physical system reacts, or transitions from one metastable state to another, is quantified by the committor function. This function contains richly detailed mechanistic information about transition pathways, but a full parameterization of the committor requires building representing a high-dimensional function, a generically challenging task. Recent efforts to leverage neural networks as a means to solve high-dimensional partial differential equations, often called "physics-informed" machine learning, have brought the committor into computational reach. Here, we build on the semigroup approach to learning the committor and assess its utility for predicting dynamical quantities such as transition rates. We show that a careful reframing of the objective function and improved adaptive sampling strategies provide highly accurate representations of the committor. Furthermore, by directly applying the Hill relation, we show that these committors provide accurate transition rates for molecular system.