Predict Protein Hydration Sites with SuperWater: Now Available on Vecura

What is SuperWater?

SuperWater is a state-of-the-art generative model designed to predict the positions of water molecules (hydration sites) on protein surfaces. By leveraging score-based diffusion and E(3)-equivariant graph neural networks (EGNNs), the model accurately models the continuous distribution of water-oxygen atoms. It offers a significant advancement over traditional computational methods by providing high-quality predictions from a single PDB structure, filling a critical gap in protein structural analysis.

What can users do with SuperWater on Vecura?

With SuperWater on Vecura, users can:

• Generate precise hydration shell predictions for protein structures, including homology models and cryo-EM data.
• Integrate predicted water positions into molecular dynamics simulations to improve starting configuration accuracy.
• Identify likely binding pocket solvation environments to support drug discovery and docking workflows.
• Visualize hydration patterns directly by merging predicted water PDB files with original protein structures in molecular viewers like PyMOL or ChimeraX.

What the output means

The output provides a water_pdb file containing the final predicted water oxygen positions, a water_positions_txt file with raw 3D coordinates for quantitative analysis, and a filter_txt file containing candidate positions with confidence scores.

This output should be used to support scientific decision making. It does not replace experimental validation.

Why this matters

The placement of water molecules is essential for understanding protein-ligand binding energetics and crystallographic refinement. However, many current structure-prediction pipelines omit these crucial components. SuperWater addresses this by enabling researchers to computationally reconstruct the solvation environment, thereby providing more biologically relevant insights into protein behavior and binding stability.

• Developed by: Research group led by Xinhao Kuang et al.
• Source: Official GitHub Repository
• Reference: Communications Chemistry (2025)