Dataset supporting the robustness of GP-based atomic energy models

Published: 3 July 2025| Version 1 | DOI: 10.17632/b75bjps2vw.1
Contributors:
Bienfait Isamura,
,
,

Description

This dataset contains both input and output files of all the experiments reported in the paper "Unprecedented robustness of physics-informed atomic energy models at and beyond room temperature" (not yet published at the date of publication of this dataset). Input files comprise (a) the reference dataset (.csv) files used for building and testing the models, and (b) CONTROL, CONFIG, and FIELD files used as inputs for FFLUX simulations. Output files include (a) model files containing the atomic energy models for each system and each mean function type, (b) the results from the distortion challenges, (c) the conformational samples from extended FFLUX simulations at 500 K, and (d) the gif files showing the prediction of restoring forces over the first 1 ps of FFLUX simulations initiated from highly distorted geometries of malondialdehyde, aspirin, and serine.

Files

Steps to reproduce

Explained in the associated paper.

Institutions

The University of Manchester

Categories

Machine Learning, Molecular Dynamics, Force Field, Gaussian Process

Funding

UK Research and Innovation

Evotec

Licence