Atomic Environment Aware Lennard-Jones Parameterization for Electrostatic Embedding ML/MM Simulations

This repository contains the supporting data, models, and figure generation code for the publication titled "Atomic Environment Aware Lennard-Jones Parameterization for Electrostatic Embedding ML/MM Simulations".

Directory Structure

• emle_models/: Contains the various EMLE models used as starting points in this study.
• datasets/: Contains the SAMPL5 dataset used for testing the EMLE models. Information about the training dataset is available in the supporting repository of the preceding paper.
• figures/: Contains notebooks used to generate the figures presented in the publication, along with the figures themselves.
• sn2_reaction: Contains the calculations and notebooks used to generate the SN2 reaction results presented in the publication.
• training/: Contains the scripts and notebooks used for training, as well as the various EMLE models trained in this study.

Associated Packages

• fes-ml: Enables hybrid ML/MM free energy calculations, with support for various alchemical modifications.
• emle-bespoke: A package which streamlines the training of EMLE models by automating conformer sampling, QM energy evaluations, and parameter fitting, with modular components for flexible use.
• PyXDM: Python package for calculating XDM (Exchange-hole Dipole Moment) multipole moments using multiple atoms-in-molecules partitioning schemes.

Citation

If you use the code, data, or models from this repository in your research, please cite the following publication:

@article{Morado_2026, 
  title={Atomic Environment Aware Lennard-Jones Parameterization for Electrostatic Embedding ML/MM Simulations}, 
  url={https://chemrxiv.org/doi/full/10.26434/chemrxiv.15004404/v1}, 
  DOI={10.26434/chemrxiv.15004404/v1}, 
  author={Morado, João and Zinovjev, Kirill and Hedges, Lester O. and Michel, Julien and Cole, Daniel J.}, 
  year={2026}, 
  month=june, 
  language={en} 
}