Using AI to model ligand binding to target proteins and affinity predictions
06 May 2026
Drug Design & Modeling
- Recent breakthroughs such as AlphaFold 3 and Boltz‑2 are expanding our ability to predict protein structures and ligand binding. Given the current accuracy and scope of these models, where are they most immediately useful in the drug discovery workflow, and where do gaps remain?
- How well do current AI-based binding and affinity prediction models handle real-world challenges such as protein flexibility, water networks, or allosteric effects, and how should these limitations be addressed?
- Looking ahead, which capabilities or improvements -- e.g., uncertainty predictions, polypharmacology predictions, generative models -- are likely to meaningfully change how AI predictions guide drug discovery in the next 3–5 years?
Industry Expert
