High-accuracy structural modeling of antibody-antigen complexes

Accurate modeling of antibody-antigen complexes is crucial for advancing therapeutics and diagnostics, yet predicting their binding interface remains a formidable challenge. To address this, we introduce DeepAAAssembly, a protocol that enhances static predicted structural information by integrating dynamic interaction patterns to guide complex assembly. Our approach leverages predicted inter-chain residue distances to construct a flexibility-aware energy function, which drives a two-stage conformational sampling process for global exploration and local exploitation. On a benchmark set of 67 representative antibody-antigen complexes, by incorporating a built-in confidence selection mechanism, DeepAAAssembly outperforms AlphaFold3, not only consistently improving the quality of most targets, achieving an average DockQ score increase of 10.7%, but also reliably raising the accuracy of the most challenging cases from incorrect to acceptable levels. Furthermore, when applied to antibody-antigen targets in the CASP16 and CASP15 benchmark sets, the average DockQ score of DeepAAAssembly showed a 16.4% improvement compared to AlphaFold3. These results demonstrate that DeepAAAssembly effectively captures conserved interaction motifs and conformational flexibility, offering a robust framework for high-accuracy antibody-antigen modeling. DeepAAAssembly enhances antibody-antigen complex modeling by integrating deep learning-based distance prediction with a global exploration and local exploitation conformational sampling strategy.