Accelerated Hydration Site Localization and Thermodynamic Profiling.

Area of Science:

• Biochemistry
• Computational Biology
• Structural Biology

Background:

• Water molecules are crucial for protein structure, function, and ligand binding.
• Accurate identification of protein hydration sites is vital for drug discovery and optimization.
• Existing computational methods for hydration site identification have limitations in accuracy and efficiency.

Purpose of the Study:

• To develop a fast and accurate computational method for localizing and thermodynamically profiling protein hydration sites (HS).
• To improve the prediction of water molecule interactions critical for ligand recognition and binding affinity.
• To facilitate the generation and optimization of lead compounds in drug discovery.

Main Methods:

• Development of a geometric deep neural network model.
• Training the model on a large dataset derived from explicit water molecular dynamics simulations.
• Validation of the model's accuracy and robustness using experimental data.

Main Results:

• The proposed method enables fast and accurate localization of protein hydration sites.
• The model provides thermodynamic profiling of identified hydration sites.
• Demonstrated utility across several case studies, confirming its practical applicability.

Conclusions:

• The novel deep learning approach significantly advances computational prediction of protein hydration sites.
• This method offers a more efficient and accurate tool for understanding water-protein interactions in drug design.
• The findings support the use of advanced computational techniques for optimizing drug affinity and selectivity.