OPRA

OPRA predicts RNA-binding sites on protein surfaces and models protein–RNA interactions by using interface propensities derived from nonredundant X-ray structures of protein–RNA complexes.

Key Features:

• Structure-derived propensities: Derives interface propensities from nonredundant X-ray structures of protein–RNA complexes.
• Propensity calculation: Calculates interface propensities between ribonucleotides and amino acid residues.
• Amino-acid preferences: Identifies arginine, lysine, and histidine as high-propensity RNA-binding residues and reports no specific preference among different ribonucleotides.
• ASA-weighted residue scoring: Assigns each protein residue a score based on its interface propensity weighted by accessible surface area (ASA).
• Patch energy computation: Computes optimal patch energy scores for each residue by aggregating the individual scores of neighboring surface residues.
• Correlation with binding sites: Shows strong correlation between computed patch scores and known RNA-binding sites on protein surfaces.
• Benchmarking: Benchmarked on a test set of 30 unbound proteins from known protein–RNA complexes with an approximate positive predictive value of 80%.
• Algorithmic prediction: Uses propensity-weighted patch scores in an algorithmic approach to predict RNA-binding areas on proteins.

Scientific Applications:

• Binding-site identification: Identification of potential RNA-binding sites on protein surfaces.
• Interaction modeling: Modeling of protein–RNA interactions for structural and functional studies.
• Molecular recognition studies: Analysis of molecular recognition mechanisms between amino acids and ribonucleotides.
• Biotechnology and therapeutics: Informing biotechnological and therapeutic research that targets protein–RNA interfaces.

Methodology:

OPRA derives interface propensities from nonredundant X‑ray protein–RNA complex structures, assigns each residue a propensity score weighted by accessible surface area (ASA), aggregates neighboring surface-residue scores to compute optimal patch energy scores, and benchmarks predictions on a test set of 30 unbound proteins.