SMILES Transformer extracts molecular fingerprints from string representations of chemical molecules.
The transformer learns latent representation that is useful for various downstream tasks through autoencoding task.
This project requires the following libraries.
- NumPy
- Pandas
- PyTorch > 1.2
- tqdm
- RDKit
Canonical SMILES of 1.7 million molecules that have no more than 100 characters from Chembl24 dataset were used.
These canonical SMILES were transformed randomly every epoch with SMILES-enumeration by E. J. Bjerrum.
After preparing the SMILES corpus for pre-training, run:
$ python pretrain_trfm.py
Pre-trained model is here.
See experiments/
for the example codes.
@article{honda2019smiles,
title={SMILES Transformer: Pre-trained Molecular Fingerprint for Low Data Drug Discovery},
author={Shion Honda and Shoi Shi and Hiroki R. Ueda},
year={2019},
eprint={1911.04738},
archivePrefix={arXiv},
primaryClass={cs.LG}
}