gr Predictor

gr Predictor predicts three-dimensional water-site distribution functions around proteins to estimate hydration structures relevant to protein folding and ligand binding.

Key Features:

• Rapid Estimation: Produces hydration-structure predictions in less than one minute on a GPU, exceeding the speed of molecular dynamics simulations and 3D-RISM by more than two orders of magnitude.
• High Accuracy: Shows a coefficient of determination of approximately 0.98 when comparing predicted distribution functions to those computed by 3D-RISM.
• Practical Application: Delivers rapid, high-fidelity hydration predictions that reduce the computational cost relative to traditional approaches.
• Integration with Existing Data: Predicted water positions align with estimates from 3D-RISM theory and with crystallographic data.

Scientific Applications:

• Protein Folding Studies: Facilitates analysis of hydration contributions to protein folding by providing detailed hydration maps.
• Ligand Binding Research: Supports investigation of water-mediated effects on ligand binding relevant to drug design and molecular interactions.

Methodology:

Leverages deep learning to estimate three-dimensional water-site distribution functions around proteins from their 3D structures, with the model trained on data derived from 3D-RISM theory.