An Integrated AI Framework for Targeted Depression Drug Discovery: Leveraging Cheminformatics and Genomics

Abstract Artificial intelligence and machine learning offer transformative opportunities for drug discovery, particularly in complex psychiatric disorders such as major depressive disorder (MDD). In this study, we introduce a dual-pipeline framework that combines cheminformatics-based compound screening with genomics-informed target discovery. We first constructed a supervised learning pipeline utilizing SMILES strings, Lipinski descriptors, and PaDEL fingerprints to predict compound activity using multiple regressors and deep learning models. In parallel, we leveraged GWAS data to identify genes implicated in depression through p-value and risk allele frequency (RAF) analyses, retrieving corresponding bioactive compounds from ChEMBL. These compounds were processed for drug-likeness and molecular descriptor profiling. Our integrated framework enhances the predictive power of AI models by anchoring chemical screening in human genetic data. The results demonstrate the potential of our approach to inform biologically grounded and data-driven discovery of novel antidepressant agents.