Comprehensive Evaluation of Fourteen Docking Programs on Protein-Peptide Complexes

Area of Science:

• Computational Biology and Cheminformatics
• Structural Bioinformatics
• Drug Discovery and Design

Background:

• Protein-peptide interactions (PpIs) are crucial for biological processes and drug design.
• Molecular docking is a key computational tool for predicting protein-peptide complex structures.
• A systematic performance assessment of docking programs for PpIs was previously lacking.

Purpose of the Study:

• To systematically evaluate the performance of various molecular docking programs for protein-peptide interactions.
• To establish a benchmark dataset (PepSet) for assessing docking program accuracy.
• To identify top-performing programs for both global and local docking of protein-peptide complexes.

Main Methods:

• Developed and utilized the PepSet benchmark dataset comprising 185 protein-peptide complexes.
• Evaluated 14 docking programs, including protein-protein, small molecule, and specialized protein-peptide docking tools.
• Introduced a new evaluation metric, IL_RMSD, incorporating the fraction of native contacts (f_nat) to measure docking accuracy.

Main Results:

• HPEPDOCK demonstrated the best performance in global docking across the PepSet dataset.
• AutoDock CrankPep (ADCP) achieved the highest accuracy in local docking predictions.
• Success rates varied significantly among the evaluated programs and docking types.

Conclusions:

• This comprehensive evaluation provides critical insights into the capabilities of current docking programs for PpIs.
• The findings guide the selection of appropriate docking tools for specific research needs.
• The study facilitates the development of more accurate and efficient docking programs for protein-peptide interactions.