ChemGAN challenge for drug discovery: can AI reproduce natural chemical diversity?

TL;DR

This paper evaluates whether AI models can generate chemically diverse molecules comparable to natural compounds, highlighting current models' limitations in reproducing true chemical diversity for drug discovery.

Contribution

It introduces a challenge to assess AI's ability to replicate natural chemical diversity and demonstrates that current models like Reinforcement Learning and ORGAN fail this test.

Findings

Current models lack sufficient chemical diversity.

Reinforcement Learning and ORGAN do not reproduce natural diversity.

The paper proposes a new challenge to stimulate future research.

Abstract

Generating molecules with desired chemical properties is important for drug discovery. The use of generative neural networks is promising for this task. However, from visual inspection, it often appears that generated samples lack diversity. In this paper, we quantify this internal chemical diversity, and we raise the following challenge: can a nontrivial AI model reproduce natural chemical diversity for desired molecules? To illustrate this question, we consider two generative models: a Reinforcement Learning model and the recently introduced ORGAN. Both fail at this challenge. We hope this challenge will stimulate research in this direction.