Junction Tree Variational Autoencoder for Molecular Graph Generation

TL;DR

This paper introduces a novel junction tree variational autoencoder that directly generates molecular graphs by constructing scaffold trees and assembling them, improving validity and performance in molecular design tasks.

Contribution

The paper presents a new graph-based generative model for molecules that outperforms prior SMILES-based methods by generating valid molecular graphs through a two-phase process.

Findings

Outperforms previous state-of-the-art models in molecular generation tasks

Maintains chemical validity at each step of molecule construction

Effective in molecule optimization and design tasks

Abstract

We seek to automate the design of molecules based on specific chemical properties. In computational terms, this task involves continuous embedding and generation of molecular graphs. Our primary contribution is the direct realization of molecular graphs, a task previously approached by generating linear SMILES strings instead of graphs. Our junction tree variational autoencoder generates molecular graphs in two phases, by first generating a tree-structured scaffold over chemical substructures, and then combining them into a molecule with a graph message passing network. This approach allows us to incrementally expand molecules while maintaining chemical validity at every step. We evaluate our model on multiple tasks ranging from molecular generation to optimization. Across these tasks, our model outperforms previous state-of-the-art baselines by a significant margin.