Water Networks in Complexes between Proteins and FDA-Approved Drugs.

Area of Science:

• Computational chemistry
• Structural biology
• Drug discovery

Background:

• Water molecules at protein-ligand interfaces are critical for drug design.
• Predicting water locations is essential when crystal structures are unavailable.
• Grand canonical Monte Carlo (GCMC) is a growing computational method for simulating water sites.

Purpose of the Study:

• To evaluate the accuracy of GCMC in predicting water binding locations.
• To assess GCMC performance using a dataset of FDA-approved drugs bound to proteins.
• To analyze factors influencing GCMC prediction accuracy and the role of water networks.

Main Methods:

• Curated a dataset of 108 protein-small molecule drug complexes.
• Applied Grand canonical Monte Carlo (GCMC) simulations.
• Analyzed prediction accuracy, sensitivity to performance metrics, and water network interactions.

Main Results:

• GCMC correctly predicted 81.4% of non-bulk water sites within 1.4 Å.
• Prediction performance is sensitive to the chosen evaluation metrics.
• Sites with more hydrogen bonds and stronger electron density were predicted more reliably.
• Over half of the structures featured ligand-contacting water networks.

Conclusions:

• GCMC is a reliable method for predicting water binding sites in drug design.
• Accurate performance evaluation requires careful consideration of metrics.
• Ligand-interacting water networks must be considered to avoid unexpected outcomes in drug design.
• Cooperative effects within water networks are important for structure-based drug design.