Predicting conserved water-mediated and polar ligand interactions in proteins using a K-nearest-neighbors genetic...

Area of Science:

• Biochemistry
• Structural Biology
• Computational Biology

Background:

• Water molecules play crucial roles in protein-ligand interactions, influencing biological processes.
• Understanding water displacement or conservation in active sites is key for drug design but not well characterized.

Purpose of the Study:

• To develop and validate a computational method (Consolv) for predicting conserved water molecules in protein active sites upon ligand binding.
• To identify environmental features that determine whether bound water is displaced or retained when a ligand binds.

Main Methods:

• Consolv utilizes a hybrid k-nearest-neighbors classifier and genetic algorithm approach.
• It analyzes the micro-environment of water molecules in free protein structures, considering features like temperature factor, hydrogen bonds, and atom density/hydrophilicity.
• The method was trained on 13 proteins and tested on seven new proteins.

Main Results:

• Consolv achieved 75% accuracy in predicting conserved active-site water molecules upon ligand binding.
• Accuracy increased to 90% when predicting conserved water or polar ligand atom binding, indicating effective assessment of polar binding sites.
• The study found that active-site water conservation is largely independent of the ligand, driven by the protein's micro-environment.

Conclusions:

• Consolv accurately predicts water molecule behavior in protein active sites, aiding ligand design by identifying optimal binding strategies.
• The tool helps in designing ligands that either displace water or utilize it for binding, and predicts water molecules likely to be displaced.
• Protein micro-environment is the dominant factor influencing water conservation or displacement, offering insights into protein-water interactions.