From Atoms to Dynamics: Learning the Committor Without Collective Variables

TL;DR

This paper presents a graph neural network that predicts the committor function directly from atomic coordinates, eliminating the need for predefined collective variables and enabling atom-level interpretability in molecular transition analysis.

Contribution

It introduces a novel GNN architecture based on geometric vector perceptrons for direct committor prediction without collective variables, enhancing interpretability and accuracy.

Findings

Accurately infers the committor function across diverse molecular systems.

Identifies key atoms involved in transition mechanisms.

Provides precise estimates of rate constants.

Abstract

This Brief Communication introduces a graph-neural-network architecture built on geometric vector perceptrons to predict the committor function directly from atomic coordinates, bypassing the need for hand-crafted collective variables (CVs). The method offers atom-level interpretability, pinpointing the key atomic players in complex transitions without relying on prior assumptions. Applied across diverse molecular systems, the method accurately infers the committor function and highlights the importance of each heavy atom in the transition mechanism. It also yields precise estimates of the rate constants for the underlying processes. The proposed approach opens new avenues for understanding and modeling complex dynamics, by enabling CV-free learning and automated identification of physically meaningful reaction coordinates of complex molecular processes.