arDock

arDock predicts protein interaction regions on protein surfaces by analyzing non-random docking patterns derived from large-scale cross-docking experiments to identify residues likely to form biological interfaces.

Key Features:

• Non-Random Docking Analysis: Uses random docking of arbitrary proteins to detect non-random aggregation of docking poses on protein surfaces indicative of biologically relevant sites.
• Surface Residue Ranking: Ranks surface residues by their propensity to form interfaces based on predefined docking experiments between the query protein and a set of arbitrary protein probes.
• Predictive Performance: Reports predictive accuracy with Area Under the Curve (AUC) values up to 0.69 using physical properties alone and up to 0.72 when combined with evolutionary information.
• Computational Configuration: Allows selection of the number of random partners and docking models to control computational cost and sampling depth.
• Multi-Interface Analysis: Identifies alternate or nonspecific interfaces in proteins with multiple partners, exemplified by PEBP (Phosphatidylethanolamine binding protein), a kinase inhibitor.

Scientific Applications:

• Biological interface prediction: Predicts likely protein-protein interaction sites on query proteins from physical docking patterns.
• Distinguishing true complexes from false positives: Helps discriminate biologically relevant docking poses from spurious docking results in docking experiments.
• Multi-interface characterization: Identifies potential alternate binding sites and nonspecific interfaces in multi-partner proteins such as PEBP.

Methodology:

Performs large-scale cross-docking of query proteins against sets of arbitrary protein probes, analyzes non-random aggregation of docking poses on protein surfaces, ranks surface residues by interface propensity from predefined docking experiments, and evaluates predictive performance using AUC, optionally combining physical properties with evolutionary information.