AI-Accelerated Discovery of Novel Compounds for GPCRs Targeted by Drugs of Abuse

PROJECT SUMMARY The addiction and overdose crisis in the United States has reached a record high, with almost 109,000 overdose deaths in 2021, the majority of which are due to opioid overdoses, according to the Centers for Disease Control and Prevention. Medications exist for treating brain Expanding the range of effective therapeutics for substance disorders caused by drug abuse, but use disorders, particularly opioid use disorders they have limitations. (OUDs), is necessary and urgent. T he National Institute on Drug Abuse's Division of Therapeutics and Medical Consequences has identified a number of G protein-coupled receptors (GPCRs) modulated by functionally distinct ligands as key targets for developing novel therapeutics to treat opioid overdose and opioid use disorders. Identifying specific and selective small-molecule ligands for these receptors is crucial for understanding their function and developing effective treatments. However, this is a challenging task due to the functional complexity of GPCRs and the difficulty in customizing ligands for them, despite advances in GPCR functional and structural biology. With the increasing availability of functional data from sophisticated bioactivity assays, billion-scale electronic chemical libraries, and ever-growing high-resolution structural information on GPCRs, it is important to develop quantitative and analytical approaches to leverage knowledge and information towards the development of effective medications for OUDs. The proposed research aims to develop Artificial Intelligence (AI)-driven strategies to design customized GPCR ligands and efficiently screen ultra-large electronic chemical libraries to speed up the discovery of novel chemical compounds with distinct pharmacological profiles targeting GPCRs linked to drug abuse. Specifically, the study will investigate the performance of deep neural network classifiers trained on large datasets of key structural and physicochemical properties of ligands and targets from receptor subfamilies and compare it with the performance of classifiers trained on features from a single GPCR of those subfamilies in distinguishing between ligands with varying efficacy at that receptor. The goal is to create a scalable platform that can add value to current rational drug design approaches for GPCRs associated with drugs of abuse by identifying lead compounds that can be developed into successful drugs for clinical applications.