Skip to content

coarse-graining/cgnet

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

cgnet

- This code is no longer being routinely maintained. For more modern implementations of CGSchNet, we recommend a Pytorch Geometric-based implemenation here:
https://github.com/torchmd/torchmd-net

Coarse graining for molecular dymamics (preprint)

Dependencies

Required:

  • numpy
  • pytorch (1.2 or higher)
  • scipy

Optional:

  • mdtraj (for cgnet.molecule only)
  • pandas (for cgnet.molecule only)
  • sklearn (for testing)
  • Jupyter (for examples)
  • matplotlib (for examples)

Usage

Clone the repository:

git clone [email protected]:coarse-graining/cgnet.git

Install any missing dependencies, and then run:

cd cgnet
python setup.py install

Notes

For compatibility with pytorch==1.1, please use the pytorch-1.1 branch. This branch currently does not include the updates for variable size and Langevin dynamics, nor some normalization options.

- CGnet models can display high variance between different training runs. For more stable models, we recommend using CGSchNet instead.

Cite

Please cite our paper in J Chem Phys:

@article{husic2020coarse,
  title={Coarse graining molecular dynamics with graph neural networks},
  author={Husic, Brooke E and Charron, Nicholas E and Lemm, Dominik and Wang, Jiang and P{\'e}rez, Adri{\`a} and Majewski, Maciej and Kr{\"a}mer, Andreas and Chen, Yaoyi and Olsson, Simon and de Fabritiis, Gianni and Noe{\'e}, Frank and Clementi, Cecilia},
  journal={The Journal of Chemical Physics},
  volume={153},
  number={19},
  pages={194101},
  year={2020},
  publisher={AIP Publishing LLC}
}

Various methods are based off the following papers. CGnet:

@article{wang2019machine,
  title={Machine learning of coarse-grained molecular dynamics force fields},
  author={Wang, Jiang and Olsson, Simon and Wehmeyer, Christoph and Pérez, Adrià and Charron, Nicholas E and de Fabritiis, Gianni and Noé, Frank and Clementi, Cecilia},
  journal={ACS Central Science},
  year={2019},
  publisher={ACS Publications},
  doi={10.1021/acscentsci.8b00913}
}

SchNet:

@article{schutt2018schnetpack,
  title={SchNetPack: A deep learning toolbox for atomistic systems},
  author={Schutt, KT and Kessel, Pan and Gastegger, Michael and Nicoli, KA and Tkatchenko, Alexandre and Müller, K-R},
  journal={Journal of Chemical Theory and Computation},
  volume={15},
  number={1},
  pages={448--455},
  year={2018},
  publisher={ACS Publications}
}