Graph Convolutional Policy Network for Goal-Directed Molecular Graph Generation

TL;DR

This paper introduces GCPN, a graph convolutional policy network that uses reinforcement learning to generate molecules optimized for specific properties while respecting chemical rules, outperforming previous methods.

Contribution

The paper presents a novel GCPN model that integrates domain-specific rules into graph generation and demonstrates significant improvements in molecular property optimization.

Findings

61% improvement in chemical property optimization

Resembles known molecules effectively

184% improvement in constrained property optimization

Abstract

Generating novel graph structures that optimize given objectives while obeying some given underlying rules is fundamental for chemistry, biology and social science research. This is especially important in the task of molecular graph generation, whose goal is to discover novel molecules with desired properties such as drug-likeness and synthetic accessibility, while obeying physical laws such as chemical valency. However, designing models to find molecules that optimize desired properties while incorporating highly complex and non-differentiable rules remains to be a challenging task. Here we propose Graph Convolutional Policy Network (GCPN), a general graph convolutional network based model for goal-directed graph generation through reinforcement learning. The model is trained to optimize domain-specific rewards and adversarial loss through policy gradient, and acts in an environment that incorporates domain-specific rules. Experimental results show that GCPN can achieve 61% improvement on chemical property optimization over state-of-the-art baselines while resembling known molecules, and achieve 184% improvement on the constrained property optimization task.