ARAE torch for SMILES generation

Prerequisite:

python3

numpy

pytorch

RDKit

SA_Score for RDKit

https://github.com/rdkit/rdkit/tree/master/Contrib/SA_Score

Download:

git clone https://github.com/gicsaw/ARAE_torch

Training data preparation :

cd ARAE_torch

generate data

python data_char.py

trainning (skip)

unconditional

python ARAE_train.py

conditional

python CARAE_train.py

generation

unconditional

python ARAE_gen.py

conditional

python CARAE_gen.py $logP $MW $QED

valid, unique, and novel check

export PYTHONPATH="SA_Score directory":${PYTHONPATH}

python valid.py result_CARAE_gen/epoch59

output file: result_CARAE_gen/epoch59/smiles_novel.txt

References:

Hong, S. H., Ryu, S., Lim, J., & Kim, W. Y. (2019). Molecular Generative Model Based On Adversarially Regularized Autoencoder. Journal of Chemical Information and Modeling.