Accurate QM/MM Molecular Dynamics for Periodic Systems in \textsc{GPU4PySCF} with Applications to Enzyme Catalysis

TL;DR

This paper introduces a GPU-accelerated QM/MM method for periodic systems, enabling accurate, energy-conserving molecular dynamics simulations with applications to enzyme catalysis.

Contribution

It presents a novel implementation of GPU-accelerated QM/MM for periodic systems with controlled errors and stable dynamics, integrated into the open-source PySCF package.

Findings

The method achieves stable, energy-conserving dynamics.

Application to chorismate mutase reveals sensitivity of catalytic rates to computational choices.

Demonstrates the effectiveness of GPU acceleration in complex biochemical simulations.

Abstract

We present an implementation of the quantum mechanics/molecular mechanics (QM/MM) method for periodic systems using GPU accelerated QM methods, a distributed multipole formulation of the electrostatics, and a pseudo-bond treatment of the QM/MM boundary. We demonstrate that our method has well-controlled errors, stable self-consistent QM convergence, and energy-conserving dynamics. We further describe an application to the catalytic kinetics of chorismate mutase. Using an accurate hybrid functional reparametrized to coupled cluster energetics, our QM/MM simulations highlight the sensitivity in the calculated rate to the choice of quantum method, quantum region selection, and local protein conformation. Our work is provided through the open-source \textsc{PySCF} package using acceleration from the \textsc{GPU4PySCF} module.