GlycoWeave models how proteins and glycans recognize each other
This new framework achieves a 71.9% Top-25 high-quality success rate on a benchmark of 698 protein-glycan complexes.
GlycoWeave is a glycan-specific diffusion framework for modeling protein-glycan recognition. It achieves a 71.9% Top-25 high-quality success rate on a benchmark of 698 protein-glycan complexes.
The framework uses chemically constrained conformational sampling and multiscale molecular representations. It also uses multi-source learning to integrate experimental complexes with structures refined using classical and quantum mechanics-informed potentials.
A heterogeneous equivariant network couples atom-level interaction geometry with monosaccharide-level structural context. This framework operates in internal-coordinate space to preserve glycan connectivity and stereochemistry.
GlycoWeave requires only 0.34 s per sampled pose. It recovers native protein-glycan interactions and reproduces linkage-specific conformational preferences.
Accurate prediction of protein-glycan complex structures is challenging. This difficulty exists because of chemical diversity, conformational flexibility, and sparse experimental structural coverage of glycans.
Key facts
- GlycoWeave achieves a 71.9% Top-25 high-quality success rate.
- The framework requires 0.34 s per sampled pose.
- GlycoWeave preserves glycan connectivity and stereochemistry.