Molecule Optimization by Explainable Evolution

@inproceedings{Chen2021MolEvol,
    title={Molecule Optimization by Explainable Evolution},
    author={Binghong Chen* and Tianzhe Wang* and Chengtao Li and Hanjun Dai and Le Song},
    booktitle={International Conference on Learning Representations},
    year={2021}
}

Overview

We release the PyTorch code for the MolEvol. [Paper]:

Details for Graph Explanation Process

Install Dependencies

You should install dependencies for chemprop and multiobj-rationale first.

Install chemprop

conda create -n chemprop python=3.8
conda activate chemprop
conda install pytorch==1.6.0 torchvision==0.7.0 -c pytorch
conda install -c conda-forge rdkit
pip install -r requirements.txt
pip install chemprop

Install multiobj-rationale

cd ..
git clone git@github.com:binghong-ml/multiobj-rationale.git
cd multiobj-rationale
pip install -e .
cd ../MolEvol

Also, please download the model files into the root directory.

How to Use

For instance, if you want to jointly optimize jnk3, gsk3, QED, SA for the molecules.

Run the following command:

export CUDA_VISIBLE_DEVICES=0
export TF_CPP_MIN_LOG_LEVEL=2
export PYTHONPATH=$PWD
python MolEvol/main.py --rounds=10 --epoch=1

You will get the molecule optimization model along with the optimized molecules in the experiment folder.