Problem
Research questions and friction points this paper is trying to address.
antibody design
geometric representation
antigen-antibody interaction
cross-molecular geometry
AbGaze: Attentive Geometric Representation Learning for End-to-End Antibody Design
🤖 AI Summary
This study addresses the challenge that existing antibody design methods struggle to capture cross-molecular geometric patterns between antigens and antibodies. We propose an end-to-end multi-task learning framework that encodes geometric features—including distances, orientations, and surface normals—within local coordinate systems via attention mechanisms. By introducing an interface context-aware adaptive aggregation strategy to construct geometric interaction representations and incorporating local frame supervision constraints, the framework enables joint training for CDR design, structure prediction, and affinity optimization. Experimental results demonstrate that our approach improves amino acid recovery by 7.1%, reduces structural error by 14.9%, enhances docking quality by 6.6%, and increases the affinity improvement rate by 32.5%.
📝 Abstract
Computational antibody design requires representations that capture the geometric patterns underlying antigen--antibody interactions, yet existing approaches often rely on scalar distances or surface-intrinsic features, leaving cross-molecular geometry largely implicit. We present AbGaze, an end-to-end antibody design framework based on attentive geometric representation learning, which encodes distance, spatial direction, and surface-normal orientation of antigen surfaces relative to antibody-residue local frames, and adaptively aggregates these geometric interactions according to their interfacial context. The learned interaction representation is shared across multi-CDR co-design, complex structure prediction, and affinity optimization, with local-frame geometric supervision further constraining the representation. AbGaze outperforms prior methods across all three tasks: relative to the second-best method, it improves amino-acid recovery by 7.1% and reduces structural error by 14.9% on average over the six CDRs, improves interface docking quality (DockQ) by 6.6%, and raises the affinity improvement rate (IMP) by 32.5%.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Attentive Geometric Representation Learning
End-to-End Antibody Design
Local-frame Encoding
Multi-CDR Co-design
Cross-molecular Geometry
🔎 Similar Papers
No similar papers found.
