Learning Multimodal Graph-to-Graph Translation for Polymer Optimization

This is an extension of the junction tree encoder-decoder model in https://arxiv.org/abs/1812.01070.

Requirements

Dependencies

• Python == 2.7
• RDKit >= 2017.09
• PyTorch >= 0.4.0
• Numpy
• scikit-learn

Big Data Tools

• Scala >= 2.11.8

The code has been tested under Python 2.7 with PyTorch 0.4.1.

Quick Start

Before you get started, you will need to create a /data folder and import a SMILES strings dataset.

You will need to split your dataset into three subsets (each separated by new line characters):

• Training pairs set of SMILES strings
• Test set of SMILES strings
• Validation set of SMILES strings

The tutorial of training and testing the variational junction tree encoder-decoder is in diff_vae/README.md.

A quick summary of different folders:

• data_prep/ contains Cosine, Jaccard, and Tanimoto similarity preprocessing scripts.
  • DataPrep_Scala/ includes Scala implementations of Cosine & Jaccard similarity.
• diff_vae/ includes the training and decoding script of variational junction tree encoder-decoder (README).
• diff_vae_gan/ includes the training and decoding script of adversarial training module (README).
• fast_jtnn/ contains the implementation of junction tree encoder-decoder.
• props/ is the property evaluation module, including penalized logP, QED and DRD2 property calculation.
• scripts/ provides model evaluation, visualization and preprocessing scripts.

Contact

Name	Email
Usman Ashraf	usman.ashraf678@gatech.edu
Rishi Gurnani	rgurnani96@gatech.edu
Kenny Scharm	kscharm3@gatech.edu
Dan Tu	dan.tu@gatech.edu