ABRefine: An Accurate Antibody Structure Refinement Method by Equivariant Graph Transformer with Rigid Body...

Area of Science:

• Computational Biology
• Structural Biology
• Protein Engineering

Background:

• Accurate antibody structure prediction is crucial for developing therapeutic antibodies.
• Current AI models like AlphaFold2/3 face challenges in precisely modeling antibody structures due to flexible complementarity-determining regions.
• Structure refinement methods are vital for enhancing the quality of predicted antibody conformations.

Purpose of the Study:

• To introduce ABRefine, an end-to-end framework for antibody structure refinement.
• To develop a method that directly predicts atomic coordinate shifts while maintaining physical consistency.
• To provide the first dedicated refinement method for nanobody structures.

Main Methods:

• Utilized an equivariant graph transformer with rigid-body constraints for structure refinement.
• Implemented a method that predicts atomic coordinate shifts without conformational sampling.
• Enforced physically meaningful rigid-body consistency during the refinement process.

Main Results:

• ABRefine demonstrates superior accuracy and efficiency compared to existing refinement methods on antibody and nanobody benchmarks.
• The framework significantly enhances the structural quality of antibody predictions, including those from AlphaFold3.
• Achieved rapid and accurate refinement of antibody structures.

Conclusions:

• ABRefine offers a novel and effective approach for antibody and nanobody structure refinement.
• The method improves the reliability of AI-driven antibody structure predictions for therapeutic applications.
• ABRefine represents a significant advancement in computational antibody design and engineering.