<ul><li>A new deep learning model called ProtAttBA has been developed for predicting binding affinity changes in antibody-antigen complexes based solely on sequence information.</li><li>ProtAttBA employs pre-training on protein sequence patterns and cross-attention-based regression to make predictions.</li><li>Evaluation on three benchmarks showed competitive performance compared to traditional methods, with notable robustness even with uncertain complex structures.</li><li>The model provides interpretability through its attention mechanism, identifying critical residues affecting binding affinity, offering a rapid and cost-effective tool for antibody engineering.</li></ul>

Sequence-Only Prediction of Binding Affinity Changes: A Robust and Interpretable Model for Antibody Engineering

Discover more