Junction Tree Variational Autoencoder for Molecular Graph Generation

This is the implementation of junction tree variational autoencoders by Jin et al. (original implementation here) ported to Python 3 and slightly improved.

Requirements

The code in this repository has only been tested on Linux. We recommend using the cross-platform miniforge Python distribution/package manager to install the dependencies (here in a conda environment called jtnn_env)

conda create --name jtnn_env --file conda_list.txt

Quick Start

The following directories contains the most up-to-date implementations:

• fast_jtnn/ contains codes for model implementation.
• fast_molvae/ contains codes for VAE training. Please refer to fast_molvae/README.md for details.

The following directories provides scripts for the experiments in our original ICML paper:

• molvae/ includes scripts for training our VAE model only. Please read molvae/README.md for training our VAE model.
• jtnn/ contains codes for model formulation.

Contact

Repository authors: Hessam Mehr (Hessam.Mehr@glasgow.ac.uk), Dario Caramelli (Dario.Caramelli@glasgow.ac.uk) Original author: Wengong Jin (wengong@csail.mit.edu)