GAME-Net-UQ

This repository contains the Python code used to train and evaluate GAME-Net-UQ, a graph neural network with uncertainty quantification (UQ) for predicting the DFT energy of relaxed species and transition states adsorbed on metal surfaces.

Install

pip install gamenet-uq

The main dependencies of the repo can be found in pyproject.toml

Dataset

The datasets used to develop GAME-Net-UQ can be found in the Zenodo repository https://doi.org/10.5281/zenodo.17977395.

1. The ASE database (212 MB) containing all relaxed DFT structures and transition states is located at gamenetuq_results/ASE_database/all.db.
2. The final, clean graph dataset (102 MB) in PyG format is located at gamenetuq_results/PyG_graph_database/all_scaled_energy_025_125_2_False_False_False_True_False. The graph dataset can be recreated from the ASE database with the gen_dataset.py script.

Model training and finetuning

To train the model, run the script train_mve.py. The input template file provides an explanation for each entry required in the training configuration file.

python train_mve.py -i input.toml -o output_dirname

Pretrained model

The final pretrained model can be employed with CARE (link).

License

The code is released under the MIT license.

Reference

Morandi, S., Loveday, O., Renningholtz, T. et al. An end-to-end framework for reactivity in heterogeneous catalysis. Nat. Chem. Eng. (2026). https://doi.org/10.1038/s44286-026-00361-8