Mastering Water Molecules: From 3D-RISM to HydraMap
Yan Li (Fudan University)
Water molecules play a crucial role in protein-ligand recognition, modulating binding thermodynamics through hydrogen bond networks and desolvation penalties. Accurate characterization of hydration sites remains a central challenge in structure-based drug design.
Existing methods, such as WATsite and 3D-RISM, provide reliable predictions through molecular dynamics or statistical mechanics, but their high computational costs and complex workflows limit their applications in high-throughput tasks. To address this issue, we developed HydraMap, an empirical knowledge-based method that predicts hydration sites using the statistical potential of atom pairs extracted from crystal structures. HydraMap achieves 1,000-fold and 30-fold speedups over WATsite and 3D-RISM, respectively, while maintaining comparable accuracy. Integrating its derived desolvation energy term (DEWED) into the X-Score function significantly enhanced screening performance on the DUD-E benchmark, with improvements in enrichment factors and AUC values for over 70% of the 102 tested targets.
In subsequent work, we extended HydraMap to include ligand-bound states by augmenting statistical potentials with solvation structures from nearly 10,000 MD-simulated small molecules. The resulting second-generation desolvation method effectively explains structure-activity relationships of multiple targets, demonstrating its practicality in lead compound optimization.
Looking ahead, we envision deep learning as the next frontier—enabling real-time, physics-grade hydration prediction through graph-based and equivariant neural networks, potentially transforming solvation-aware scoring into a routine component of drug discovery pipelines.