InterPep2

InterPep2 predicts peptide-protein interactions and models bound peptide conformations to identify binding sites and orientations, with emphasis on interactions involving intrinsically disordered regions.

Key Features:

• Template-Based Prediction: Uses structural templates derived from protein-protein interactions to predict peptide-protein binding sites.
• Docking Capability: Docks peptides onto predicted binding surfaces to model spatial orientation of the peptide-protein complex.
• Scoring Mechanism: Employs a RandomForest algorithm trained to predict DockQ-scores from sequence and structural features to rank interaction templates.
• Performance Evaluation: Validated on a dataset of 251 peptide-protein complexes, correctly positioning peptides at the binding site in 54% of cases versus 45% for the second-best method.
• Confidence Scoring: Achieves 59% recall at a 10% false positive rate compared to 44% for the closest competitor.

Scientific Applications:

• Protein-peptide interaction studies: Predicts and models peptide binding to proteins to investigate interaction mechanisms.
• Intrinsically disordered region analysis: Targets interactions involving intrinsically disordered regions to elucidate their structural roles.
• Gene regulation and cell cycle research: Provides structural hypotheses for peptide-mediated regulation relevant to gene expression and cell cycle control.
• Drug discovery and molecular biology: Supplies modeled peptide-protein complexes and scores useful for structure-based investigation in drug discovery and bioinformatics workflows.

Methodology:

Applies a template-based approach using templates from protein-protein interactions, docks peptides onto predicted binding surfaces, and uses a RandomForest trained on sequence and structural features to predict DockQ-scores.