ChemRefine: An open-source automated and interoperable platform for machine learning and quantum chemistry simulations

ChemRxiv

Abstract

The role of machine learning (ML) in computational chemistry is growing rapidly, promising to expedite many routine tasks while improving prediction accuracy. However, the interface of newly developed machine-learned interatomic potentials (MLIPs) with existing computational chemistry workflows based on quantum mechanics (QM) remains a challenge. Here, we introduce ChemRefine as a modular automated computational chemistry platform that enables researchers to seamlessly integrate the next generation of MLIPs into their computational workflows. We road-test ChemRefine on multiple computational tasks commonly encountered by researchers, including conformational sampling, host–guest binding, redox and photophysical property prediction, transition state finding, molecular dynamics spawning, and the creation of optimal microsolvation ensembles. We also introduce a feature that enables automatic training/fine-tuning and deployment of MLIPs during workflows, enabling computational chemists to accelerate their simulations while maintaining prediction accuracy. To simplify the use of ChemRefine, we also introduce a custom large language model, ChemRefineGPT, that generates all required input and configuration files based on a natural language description of the desired workflow.

Authors 2

1. Ignacio Migliaro

   The University of Texas at Dallas

   Affiliation as printed

   The University of Texas at Dallas

2. Alistair J. Sterling

   The University of Texas at Dallas

   Affiliation as printed

   The University of Texas at Dallas