Incorporating Water Molecules into Highly Accurate Binding Affinity Prediction for Proteins and Ligands

Area of Science:

• Computational chemistry
• Structural biology
• Drug discovery

Background:

• Protein-ligand interactions are crucial in biological processes.
• Accurate prediction of binding affinity is essential for drug development.
• Existing methods often neglect the role of water molecules in binding.

Purpose of the Study:

• To develop a novel computational model, GraphWater-Net, for predicting protein-ligand binding affinity.
• To investigate the impact of incorporating water molecules into binding affinity prediction models.
• To enhance the accuracy of binding affinity predictions for rational drug design.

Main Methods:

• GraphWater-Net utilizes topological structures to represent proteins, ligands, and water molecules.
• The model employs the Graphormer network to extract interaction features.
• Attention weights and Softmax function are used for regression prediction of binding affinity.

Main Results:

• GraphWater-Net significantly improves protein-ligand binding affinity prediction performance.
• The inclusion of water molecules enhances the model's accuracy compared to methods that omit them.
• Experimental results on the CASF 2016 test set show improved Pearson correlation coefficient (R) values.

Conclusions:

• Incorporating water molecules into binding affinity prediction models is crucial for accuracy.
• GraphWater-Net offers a promising approach for predicting protein-ligand binding affinity.
• The model has the potential to accelerate rational drug design and discovery efforts.