Research work · 2026
Machine Learning in Ligand-based Drug Design
Nikita Krishna, Mohammad Imran Siddiqi
Machine Learning and Deep Learning in Drug Design
Abstract
The evolution of machine learning (ML) has fundamentally transformed ligand-based drug design (LBDD) by enabling the rapid identification, optimization, and prediction of bioactive compounds without reference to target structural information. This chapter investigates in great detail how different algorithms—from supervised ML models like random forests and support vector machines to deep learning and reinforcement learning frameworks—are used to examine vast chemical environments and predict ligand–target interactions in LBDD. Molecular representation strategies, including descriptors, fingerprints, and graph-based forms, are addressed here for the basis of developing predictive models. By presenting representative case studies, this chapter highlights successful ML-driven discoveries in many therapeutic domains, including COVID-19, diabetes, and cancer. It also discusses the shortcomings of current ML systems, including data scarcity, model interpretability, repeatability, and fit into current experimental techniques. To overcome the limitations, the chapter suggests increasing model transparency and clinical value by interdisciplinary collaboration, federated learning, and explainable artificial intelligence (XAI). The chapter concludes with a discussion on ethical frameworks and regulatory cooperation to guarantee the safe and efficient adoption of ML in the next drug discovery pipelines.