PepSet

PepSet provides a benchmark dataset of 185 protein-peptide complexes (peptide length 5–20 residues) for evaluating protein-peptide docking programs and quantifying docking accuracy using IL_RMSD and the fraction of native contacts (f_nat).

Key Features:

• Dataset composition: 185 experimentally determined protein-peptide complexes with peptide lengths ranging from 5 to 20 residues.
• Docking programs evaluated: Comparative evaluation includes 14 programs: protein-protein docking (ZDOCK, FRODOCK, HawkDock), small-molecule docking (GOLD, Surflex-Dock, AutoDock Vina), and protein-peptide docking (GalaxyPepDock, MDockPeP, HPEPDOCK, CABS-dock, pepATTRACT, DINC, AutoDock CrankPep (ADCP), HADDOCK peptide docking).
• IL_RMSD metric: IL_RMSD is introduced to measure docking accuracy by considering the fraction of native contacts (f_nat).
• Global docking performance: Reported success rates for HPEPDOCK in global docking are 4.3% (top 1), 24.3% (top 10), and 55.7% (top 100).
• Local docking performance: Reported success rates for AutoDock CrankPep (ADCP) in local docking are 11.9% (top 1), 37.3% (top 10), and 70.3% (top 100).

Scientific Applications:

• Benchmarking docking algorithms: Enable systematic comparison of protein-protein, small-molecule, and protein-peptide docking methodologies on a common dataset.
• Assessing docking accuracy metrics: Support evaluation of IL_RMSD and fraction of native contacts (f_nat) as measures of prediction quality.
• Informing peptide and peptidomimetic design: Provide data to guide design efforts aimed at modulating protein–protein interactions (PPIs).
• Method development: Serve as a standardized dataset for development and validation of improved protein-peptide docking programs.

Methodology:

Fourteen docking programs were run on the 185 complexes and predictions were assessed using IL_RMSD and fraction of native contacts (f_nat), with success rates reported at the top 1, 10, and 100 prediction levels.