SMICLR: Contrastive Learning on Multiple Molecular Representations for Semisupervised and Unsupervised Representation Learning

Authors: Gabriel A. Pinheiro, Juarez L. F. Da Silva, Marcos G. Quiles

Link to Paper

SMICLR

The official PyTorch implementation of SMICLR: Contrastive Learning on Multiple Molecular Representations for Semisupervised and Unsupervised Representation Learning. SMICLR jointly trains a graph encoder and an encoder for the simplified molecular-input line-entry system (SMILES) string to perform the contrastive learning objective.

Dependencies

We conducted the experiments using the libraries version described in the requirements.txt file.

Cite

Please cite our paper if you use this code in your own work:

@article{Pinheiro_2022_SMICLR,
  author = {Pinheiro, Gabriel A. and Da Silva, Juarez L. F. and Quiles, Marcos G.},
  title = {SMICLR: Contrastive Learning on Multiple Molecular Representations for Semisupervised and Unsupervised Representation Learning},
  journal = {Journal of Chemical Information and Modeling},
  volume = {62},
  number = {17},
  pages = {3948-3960},
  year = {2022},
  doi = {10.1021/acs.jcim.2c00521},
  note = {PMID: 36044610},
  URL = {https://doi.org/10.1021/acs.jcim.2c00521},
  eprint = {https://doi.org/10.1021/acs.jcim.2c00521}
}