Structural plausibility without binding specificity: limits of AI antibody-antigen prediction

Our next session of High-Affinity Talks features Eva Smorodina, PhD Fellow and Computational Structural Biologist at the University of Oslo, who presents new insights on antibody-antigen structure prediction. ​Antibody-antigen binding prediction remains a central challenge for AI-driven therapeutic discovery, particularly in discriminating cognate interactions from structurally plausible but incorrect pairings. ​This seminar presents a controlled, AI-method- and antibody-format-agnostic evaluation framework that measures binding specificity under realistic conditions. Using 106 experimentally determined single-chain antibody (nanobody)-antigen complexes and 11,130 shuffled non-cognate pairings, the team benchmarked publicly-available state-of-the-art structure prediction methods (AlphaFold3, Boltz-2, Chai-1). ​Although the methods tested often generated geometrically plausible complexes, internal confidence metrics (ipTM) frequently failed to discriminate correct from incorrect pairings. Increased sampling improved structural refinement but not pairing discrimination, indicating that computational resources are better allocated across independent seeds and explicit negative controls. ​To conclude, internal confidence scores are not inherently calibrated to binding specificity and require validation against realistic decoys. To enable community benchmarking and method development, the team has released ∼1.8 million AI-generated complex structures and guidance for the benchmarks ahead.