MACE-OpenMM Hackathon

This repo contains work from the MACE-OpenMM hackathon hosted by the cole-group at the University of Newcastle 2022.

This work builds upon the framework laid out by the Chodera lab in the qmlify package and accompanying paper.

AIMS

• ✅ Build an openmm-ml interface for the MACE. The interface can be found here.
• ✅ Run hybrid MM/ML simulations of ligands in complex and solvent.
• ✅ Develop a general interface built on OpenMM tools to calculate MM/ML free energy corrections starting from the output of MM relative binding calculations.
• ✅ Implement a QCEngine interface to MACE allowing for single point calculations, geometry optimisations and torsiondrives. An example notebook using the interface to perform a geometry optimisation is also included.

A conda environment file is also supplied which was used to run the hybrid MM/ML simulations. This should be used to create the fresh environment using conda env create -f env.yaml this following packages should then be manually installed

• openmm-ml
• mace
• openmmtools (Dominic's version)

Authors

• Kovács Dávid Péter
• Finlay Clark
• Harry Moore
• Mateusz Bieniek
• Josh Horton
• Daniel Cole

Acknowledgements

• Dominic Rufa
• John Chodera