AR-Pred

AR-Pred predicts active and regulatory (allosteric) binding sites in proteins to identify functional and modulatory residues for understanding protein function and supporting drug discovery.

Key Features:

• Data integration: Combines protein geometry, evolutionary data, physicochemical properties, and dynamic information to predict potential active and allosteric site residues.
• Protein dynamics: Incorporates intrinsic protein dynamic information as a predictor for binding site identification.
• Machine learning ensemble: Employs random forest algorithms trained on multiple balanced datasets to produce an ensemble of discrete models tailored for active site and allosteric site prediction.

Scientific Applications:

• Protein function analysis: Identifies active and allosteric sites to aid elucidation of protein functional mechanisms.
• Drug discovery: Predicts regulatory (allosteric) binding sites for applications in drug design targeting protein modulation.
• Comparative evaluation: Enables benchmarking against existing methods using independent protein sets to assess prediction performance.

Methodology:

Integrates protein geometry, evolutionary, physicochemical and dynamic features; uses random forest algorithms to build an ensemble of discrete models trained and validated on multiple balanced datasets for active and allosteric site prediction; reported median performance metrics are AUC 91% and MCC 0.68 for active site prediction, and AUC 80% and MCC 0.48 for allosteric site prediction.