Scoring of protein-protein docking models utilizing predicted interface residues.

Area of Science:

• Computational biology
• Structural bioinformatics
• Drug discovery

Background:

• Scoring molecular docking solutions is challenging due to vast numbers of generated models.
• Current scoring functions have limitations in accurately ranking acceptable docking poses.
• Accurate identification of protein-protein interaction interfaces is crucial for understanding biological processes.

Purpose of the Study:

• To evaluate the utility of predicted interface residues for scoring docking models.
• To benchmark various interface prediction methods in the context of molecular docking.
• To assess the impact of interface prediction quality on the success rate of docking protocols.

Main Methods:

• Systematic testing of diverse interface prediction algorithms.
• Application of prediction methods to score over 300,000 low-resolution docking decoys.
• Benchmarking against established molecular docking algorithms.

Main Results:

• Contact-based interface prediction using BIPSPI demonstrated superior performance in scoring docking solutions.
• Over 12% of first-ranked docking models were identified as acceptable using the BIPSPI method.
• Precision emerged as a key metric for evaluating interface prediction quality for docking constraints.

Conclusions:

• Predicted interface residues offer a viable strategy for improving molecular docking model scoring.
• BIPSPI's contact-based approach is effective for ranking docking solutions.
• Further research is needed to overcome limitations in integrating interface predictions into docking protocols.