Artificial intelligence-assisted lead optimization in drug discovery: bridging computational advances and translational challenges.

Abstract

Lead optimization is a resource-intensive stage of drug discovery requiring the simultaneous optimization of potency, selectivity, pharmacokinetics, safety, and synthetic feasibility. Although artificial intelligence (AI) and machine learning (ML) have substantially advanced predictive modeling, virtual screening, de novo molecular design, and multiparameter optimization, their translation into routine medicinal chemistry remains challenging. This review critically examines current AI approaches for lead optimization, emphasizing the gap between computational benchmark performance and practical medicinal chemistry applications. Recent advances in graph neural networks (GNNs), transformer architectures, diffusion models, and chemical foundation models have expanded AI-assisted molecular design and property prediction. We further discuss emerging concepts including data-centric AI, uncertainty quantification, trustworthy AI, the AI optimization paradox, and the shift from molecular prediction toward scientific decision-making as key determinants of successful implementation. Despite increasing industrial adoption, AI remains dependent on high-quality experimental data, model generalizability, and rigorous experimental validation. We conclude that future progress will depend less on increasingly sophisticated algorithms than on trustworthy AI systems that improve scientific decision-making within iterative lead optimization workflows and ultimately enhance translational success in drug discovery.