3D MOLECULAR DOCKING COMBINED WITH AI-DRIVEN VIRTUAL SCREENING TO REPURPOSE FDA-APPROVED DRUGS FOR RARE NEURODEGENERATIVE DISEASES

Abstract

Drug repurposing presents an efficient and cost-effective pathway for accelerating therapeutic discovery, particularly for rare neurodegenerative diseases where conventional drug development is economically prohibitive and time-intensive. This study proposed a hybrid computational framework combining molecular docking, artificial intelligence (AI)-driven virtual screening, and machine learning classification to identify new therapeutic applications for FDA-approved compounds. Over 200 drugs were screened against validated disease targets using 3D molecular docking, and 20 top candidates were selected based on binding energy thresholds (ΔG < -7.0 kcal/mol). A Random Forest classifier trained on molecular descriptors predicted ADMET properties with 91% accuracy, while SHAP analysis highlighted key features influencing model performance. Further validation using 100 ns molecular dynamics simulations confirmed complex stability for high-ranking compounds. Deep learning models, including graph neural networks, predicted biological activity and toxicity with strong alignment to docking results. Literature-based triangulation identified five candidates with high therapeutic promise and prior pharmacological evidence. The results, visualized through comprehensive tables and 12 analytical plots, emphasize the effectiveness of this integrative pipeline in narrowing down repurposing candidates with both structural viability and clinical relevance. This study provides a replicable, scalable, and intelligent methodology to streamline the drug discovery process, especially for underfunded and neglected disease domains. The proposed pipeline significantly reduces experimental burden while maximizing translational impact, demonstrating that AI-guided drug repurposing is a transformative strategy in modern pharmaceutical research.