Reactron

Implementation of electron movements prediction with Reactron developed by Yousung Jung group's at Seoul National University.

Contents

• Developer
• OS Requirements
• Hardware Requirements
• Python Dependencies
• Installation Guide
• Reproducing results
• Citation
• License

Developer

Shuan Chen (shuan.micc@gmail.com) and Kye Sung Park (kspark.micc@gmail.com)

OS Requirements

This repository is only reproducible on Linux operating systems. The compiled binary modules (.so files) are built for Linux x86_64 with Python 3.8.

Hardware Requirements

This repository has been tested on NVIDIA GeForce RTX 3090 GPU, but other CUDA-based GPU should also work at the right PyTorch version.

Python Dependencies

• Python (version == 3.8)
• Numpy (version >= 1.16.4)
• PyTorch (version >= 1.0.0)
• RDKit (version >= 2019)
• DGL (version >= 0.5.2)
• DGLLife (version >= 0.2.6)
• Pandas
• Scikit-learn
• tqdm

Installation Guide

Create a virtual environment to run the code of Reactron.
Make sure to install pytorch with the cuda version that fits your device.
This process usually takes a few minutes to complete.

git clone https://github.com/snu-micc/Reactron.git
cd Reactron
conda create -c conda-forge -n rdenv  python=3.6 -y
conda activate rdenv
conda install pytorch cudatoolkit=11.3 -c pytorch -y
conda install -c conda-forge rdkit -y
conda install -c dglteam dgl-cuda11.3
pip install dgllife

Reproducing results

See Demo.ipynb for running instructions and expected output. The instruction of how to reproduce the reactions shown in the paper is given in the same notebook. The prediction usually takes no more than 1 second for each reaction on an NVIDIA GeForce RTX 3090 GPU.

Citation

under review

License

This project is covered under the CC BY-NC-SA 4.0 (Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International) license. See the LICENSE file for details.