Conditional Constrained Graph Variational Autoencoders for Molecule Design

TL;DR

This paper introduces CCGVAE, a graph variational autoencoder that incorporates atom valence histograms to improve molecule generation, achieving better results on standard datasets.

Contribution

It proposes a novel conditional constrained graph VAE that leverages chemical valence information to enhance molecule generation quality.

Findings

Improved evaluation metrics on benchmark datasets.

Effective incorporation of atom valence histograms.

State-of-the-art performance in molecule design.

Abstract

In recent years, deep generative models for graphs have been used to generate new molecules. These models have produced good results, leading to several proposals in the literature. However, these models may have troubles learning some of the complex laws governing the chemical world. In this work, we explore the usage of the histogram of atom valences to drive the generation of molecules in such models. We present Conditional Constrained Graph Variational Autoencoder (CCGVAE), a model that implements this key-idea in a state-of-the-art model, and shows improved results on several evaluation metrics on two commonly adopted datasets for molecule generation.