Abstract
Predicting antigen-antibody binding is essential to drug discovery and protein engineering. For de novo antibody design, generalizable binding prediction models are crucial for efficient in silico screening. However, existing affinity predictors lack generalization, with performance deteriorating for antibodies targeting antigens absent from training data or datasets lacking non-binders. To address this, we establish a benchmarking framework for evaluating universal antibody-antigen binding affinity prediction. Our framework compares sequence- and structure-based methods across diverse antigens, introducing standardized evaluation protocols based on pairwise accuracy and retrieval metrics. We propose MochiBind, a sequence-only pairwise binding affinity predictor, and benchmark it against structure-derived baselines such as Boltz-2, GeoDock, and Graphinity. The results show that MochiBind achieves comparable or superior performance in pairwise accuracy and retrieval, suggesting that sequence-based approaches can match or surpass structure-based models in generalization. The proposed benchmark provides a foundation for fair comparison and future development, enabling scalable, sequence-driven solutions to binding affinity prediction.
[Display omitted]
•Ranking-based evaluation framework for binding affinity prediction models•Large antibody pool per antigen system enables robust out-of-sample testing•Sequence-based model with pairwise training generalizes to held-out antigens•MochiBind outperforms existing benchmark models in ranking-based evaluation
Immunology; Structural biology; Bioinformatics