Artificial Intelligence and Machine Learning in Drug Discovery and Pharmacology: Applications, Explainability, and Future Directions

Abstract

Drug discovery and pharmacology are undergoing a significant transformation driven by advances in artificial intelligence (AI) and machine learning (ML). What historically required a decade or longer and substantial capital investment per approved therapeutic is increasingly being reshaped by data-driven computational pipelines capable of accelerating target identification, virtual screening, lead optimization, toxicity prediction, and drug repurposing. This review synthesizes recent developments in AI/ML applications across the drug discovery and development pipeline, with particular attention to deep learning architectures such as graph neural networks and transformer-based models, generative molecular design methods, and reinforcement-learning strategies for lead optimization. It further examines the growing importance of explainable AI (XAI) in pharmacology, where regulators and clinicians increasingly demand transparent, auditable model reasoning rather than opaque predictions. Persistent challenges — including data quality and bias, limited generalizability across chemical space, the interpretability-performance trade-off, and evolving regulatory expectations — are discussed alongside emerging directions such as federated learning, multimodal foundation models, and physics-informed explainability. The review argues that the next phase of AI-enabled pharmacology will be defined less by predictive accuracy alone and more by the trustworthiness, reproducibility, and clinical integration of these systems.