MACHINE LEARNING–ENHANCED MOLECULAR DOCKING AND VIRTUAL SCREENING FOR DRUG REPURPOSING IN INFLAMMATORY DISEASES

Abstract

Inflammatory diseases represent a broad class of acute and chronic conditions characterized by dysregulated immune responses and persistent tissue damage, posing substantial clinical and socioeconomic burdens worldwide. Drug repurposing has emerged as a time- and cost-efficient strategy to identify new therapeutic indications for existing compounds with established safety profiles. In this context, machine learning–enhanced molecular docking and virtual screening have gained increasing prominence as integrative computational approaches capable of accelerating the identification of candidate drugs targeting inflammation-associated biomolecules. This study presents a comprehensive framework that combines classical structure-based molecular docking with supervised and deep learning models to improve binding affinity prediction, ranking accuracy, and hit enrichment during virtual screening campaigns. Molecular descriptors, protein–ligand interaction fingerprints, and docking-derived scores are leveraged as input features to train predictive models that discriminate high-affinity binders from inactive compounds across multiple inflammatory targets. The proposed workflow is applied to curated drug libraries and key protein targets implicated in inflammatory signaling pathways, enabling the prioritization of repurposable compounds with favorable interaction profiles. Results demonstrate that machine learning integration significantly enhances screening performance compared to conventional docking-only strategies by reducing false positives and improving early recognition of promising candidates. This approach underscores the potential of hybrid computational methodologies to support rational drug repurposing in inflammatory diseases and provides a scalable platform for translational pharmacological research.