Datasets and Trained Machine Learning Models for Potential Energy Surfaces of Hydrogen Atom Transfer in Peptides

Datasets and trained machine learning models for the paper: 'Learning Potential Energy Surfaces of Hydrogen Atom Transfer Reactions in Peptides' (Neubert, Gräter, Friederich, 2025). It includes DFT (bmk/def2-TZVPD) and semiempirical (xTB) configurations, along with corresponding energies and forces for a diverse set of hydrogen atom transfer reactions in peptide environments. Trained graph neural network models (SchNet, Allegro, and MACE) are included, which have been trained on the full dataset and subsets.