Generative network complex for the automated generation of druglike molecules

TL;DR

This paper introduces a generative network complex that creates novel drug-like molecules by optimizing multiple properties in the latent space of an autoencoder, aiding drug discovery especially for low-income populations.

Contribution

It presents a new generative framework combining multi-property optimization and independent validation for efficient drug-like molecule generation.

Findings

Generated hundreds of new drug candidates including BACE1 inhibitors.

Successfully created alternative candidates for eight market drugs.

Validated molecule predictions with independent 2D fingerprint-based screening.

Abstract

Current drug discovery is expensive and time-consuming. It remains a challenging task to create a wide variety of novel compounds with desirable pharmacological properties and cheaply available to low-income people. In this work, we develop a generative network complex (GNC) to generate new drug-like molecules based on the multi-property optimization via the gradient descent in the latent space of an autoencoder. In our GNC, both multiple chemical properties and similarity scores are optimized to generate and predict drug-like molecules with desired chemical properties. To further validate the reliability of the predictions, these molecules are reevaluated and screened by independent 2D fingerprint-based predictors to come up with a few hundreds of new drug candidates. As a demonstration, we apply our GNC to generate a large number of new BACE1 inhibitors, as well as thousands of novel alternative drug candidates for eight existing market drugs, including Ceritinib, Ribociclib, Acalabrutinib, Idelalisib, Dabrafenib, Macimorelin, Enzalutamide, and Panobinostat.