New benchmark metrics for protein-protein docking methods.

Area of Science:

• Computational Biology
• Structural Biology
• Biophysics

Background:

• Accurate prediction of protein-protein complex structures is crucial for understanding biological processes.
• Existing methods for assessing the quality of protein complex models have limitations in detecting subtle similarities and their statistical significance.

Purpose of the Study:

• To develop and validate novel scoring functions for benchmarking the performance of computational methods predicting protein-protein complex structures.
• To introduce the Interface Similarity score (IS-score) as a more suitable metric for assessing protein docking models.

Main Methods:

• Introduction of two scoring functions: the interfacial Template Modeling score (iTM-score) for geometric interface assessment and the Interface Similarity score (IS-score) for combined geometric and residue contact similarity.
• Validation of the IS-score using a large-scale benchmark of 1562 dimeric complexes.
• Application of the IS-score to evaluate models from the Critical Assessment of Prediction of Interactions (CAPRI) experiments.

Main Results:

• The IS-score demonstrates superior suitability for assessing protein docking models compared to the iTM-score.
• Large-scale benchmark testing confirms the robustness and effectiveness of the IS-score.
• Application to CAPRI data shows general consistency with existing assessments while highlighting previously underestimated model significance.

Conclusions:

• The IS-score provides a more comprehensive and accurate evaluation of protein-protein complex models.
• This scoring function can refine the assessment of computational docking predictions, particularly in the context of challenges like CAPRI.
• The IS-score aids in identifying high-quality models that might be overlooked by other metrics.