/ANA2B

Primary LanguageJupyter NotebookMIT LicenseMIT

ANA2B

Repository for the ANA2B publication.

Reproduction / Examples

Weights for the models necessary to reproduce results can be found in source/weights_models/

Necessary datasets (need to be downloaded separately):

For Training:

For Validation and Testing:

Examples and scripts necessary to reproduce results can be found in

  • MD: source/anaff/md
  • CSP: source/anaff/crystal_ranking
  • Dimers: source/results
  • ANA2B model training: source/anaff/train_model_alpha_gnn.py
  • A small example for general usage: source/examples/example.ipynb

Installation Instructions

After cloning the repo, following package are required:

  • Tensorflow
  • ase
  • mdtraj
  • graph_nets
  • Simple DFT-D3

Exemplary usecases can be found in

  • source/examples/example.ipynb
  • source/anaff/md/run_md_ase.py

Citation

@article{ANA2B
      author ="Thürlemann, Moritz and Riniker, Sereina",
      title  ="Hybrid classical/machine-learning force fields for the accurate description of molecular condensed-phase systems",
      journal  ="Chem. Sci.",
      year  ="2023",
      doi  ="10.1039/D3SC04317G",
}

License

MIT