BarclayII / icml18-jtnn

Junction Tree Variational Autoencoder for Molecular Graph Generation

Implementation of Junction Tree Variational Autoencoder from Jin et al. https://arxiv.org/abs/1802.04364

This is a fork to port the model onto Python 3 and some library.

Currently only VAE training is ported.

You need to initialize the submodule by the following:

git submodule init
git submodule update

Checklist

• VAE training
  • Run PYTHONPATH=$PWD/_dgl/python:$PWD:$PYTHONPATH python3 molvae/vaetrain.py -t data/train.txt -v data/vocab.txt -s . -b 10
  • Or PYTHONPATH=$PWD/_dgl/python:$PWD:$PYTHONPATH python3 molvae/vaetrain_dgl.py -t data/train.txt -v data/vocab.txt -s . -b 10
• VAE decoding during inference
• Property prediction
• CUDA support
• Profiling
  • To enable profiling, set the environment variable PROFILE to 1, and set PROFILE_OUTPUT to the destination of line profiler dump (leave it blank for standard output)

Requirements

• Linux (We only tested on Ubuntu)
• RDKit (version >= 2017.09)
• Python (version >= 2.7)
• PyTorch (version >= 0.2)

To install RDKit, please follow the instructions here http://www.rdkit.org/docs/Install.html

We highly recommend you to use conda for package management.

Quick Start

This repository contains the following directories:

• bo/ includes scripts for Bayesian optimization experiments. Please read bo/README.md for details.
• molvae/ includes scripts for training our VAE model only. Please read molvae/README.md for training our VAE model.
• molopt/ includes scripts for jointly training our VAE and property predictors. Please read molopt/README.md for details.
• jtnn/ contains codes for model formulation.

Contact

Quan (Andy) Gan ([e-mail])