Data for: Incorporating Non-Covalent Interactions in Transfer Learning Gaussian Process Regression Models for Molecular Simulations

This repository provides additional data to accompany the paper:

"Incorporating Noncovalent Interactions in Transfer Learning Gaussian Process Regression Models for Molecular Simulations"
M. L. Brown, B. K. Isamura, J. M. Skelton and P. L. A. Popelier
Journal of Chemical Theory and Computation (2024)
DOI: 10.1021/acs.jctc.4c00402

This article utilises s a dimeric GPR model in the quantum chemical topology based force field, FFLUX, for the first time, allowing for simulations free from non-bonded potentials. This model is benchmarked against a monomer model (combined with a Lennard-Jones potential). Transfer learning is also implemented to counteract the cost of training systems with increased dimensionality. This repository makes available a full set of data from these calculations, including:

• Energies of dimers used to calculate formation energies;
• Gaussian process regression (GPR) models used in simulations;
• Data for the calculated IR spectra;
• Files used to generate the optimised structures;
• Visual supporting data for the paper;
• Data used to train and test models;
• Forces used for vibrational frequency calculations.

For details of how this data was generated, users are referred to the published article and supporting information.

While DL_FFLUX (the program that houses the FFLUX force field) has not yet been made publicly available, input files and the GPR models are provided for when it is made available.