Generate 3D transition state geometries with GNNs (Note: python3 pytorch version and integration into ARC coming soon!)
- python (version=2.7)
- tensorflow (version=1.14)
- rdkit (version=2018.09.3)
git clone https://github.com/PattanaikL/ts_gen
To train the model, call the train.py script with the following parameters defined. If training with your own data, ensure data is in sdf format and molecules between reactants, products, and transition states are all aligned.
python train.py --r_file data/intra_rxns_reactants.sdf --p_file data/intra_rxns_products.sdf --ts_file data/intra_rxns_ts.sdf
To evaluate the trained model, refer to use_trained_model.ipynb
Currently, we support sdf integration through rdkit, but all that's required is an rdkit mol. If you have data in xyz format, consider using the code from the Jensen group to convert to rdkit.