Water-guided docking improves prediction of protein–glycan complexes

Accurate prediction of structure of protein-carbohydrate complexes remains a significant challenge in structural glycobiology, largely due to the flexibility of glycans and the shallow, hydrophilic nature of their binding sites. To address this issue, we developed a guided docking protocol that leverages Crystallographic Water Sites (CWS) to enhance glycan pose prediction using AutoDock Vina (ADV). By defining Waters Ideal Interactions (WII)-interaction hotspots derived from water molecules in apo structures-the protocol systematically rewards chemically meaningful receptor-ligand contacts during docking simulations. The WII Guided Approach (WIIGA) was benchmarked against a curated dataset of 30 high-quality protein-oligosaccharide complexes, which included ligands ranging from tetra- to nonasaccharides. Performance evaluation demonstrated that the guided protocol consistently outperformed conventional methods (ADV, Vina Carb (VC), Vina Carb with CH-π (VC CH-π) and GlycoTorch Vina (GTV)), delivering improved pose prediction accuracy. Our method proved robust even in the absence of holo structures and was effective in cross-docking drug-like glycomimetics. The protocol is easy to implement and broadly applicable to a wide range of glycan-binding proteins. These findings underscore the value of solvent-derived information for improving docking accuracy and support the use of guided approaches as a versatile tool for glyco-ligand modeling and structure-based design.