AI-driven innovation in antibody-drug conjugate design

Abstract

Antibody-drug conjugates (ADCs) represent a mechanistically defined class of targeted therapeutics that combine monoclonal antibodies with cytotoxic payloads to achieve selective delivery to antigen-expressing carcinoma cells. Conventional ADC development has primarily relied on empirical screening and structure-based design, often limited by incomplete structural information, non-systematic linker–payload selection, and constraints in experimental throughput. Computational methods, including artificial intelligence and machine learning (AI/ML) are increasingly being integrated into ADC discovery and optimization workflows (i.e., AI-driven ADC Design) to address these limitations. This review is organized into six sections: (1) the progression from traditional modeling approaches to AI-driven design of individual ADC components; (2) the application of deep learning (DL) to antibody structure prediction and identification of optimal conjugation sites; (3) the use of AI/ML models for forecasting pharmacokinetic properties and toxicity profiles; (4) emerging generative algorithms for antibody sequence diversification and affinity optimization; (5) case studies demonstrating the integration of computational tools with experimental pipelines, including systems that link in silico predictions to high-throughput validation; and (6) persistent challenges, including data sparsity, model interpretability, validation complexity, and regulatory considerations. The review concludes with a discussion of future directions, emphasizing the role of multimodal data integration, reinforcement learning (RL), and closed-loop design frameworks to support iterative ADC development.

Topics & Concepts