hybridPBRpred

hybridPBRpred predicts protein-binding residues from protein sequence by combining SCRIBER and disoRDPbind to provide comprehensive PBR identification across structured and intrinsically disordered proteins.

Key Features:

• Integration of Predictive Models: Combines SCRIBER, trained on structured protein-protein complexes, with disoRDPbind, trained on disordered protein-binding residues, to provide complementary sequence-based predictions.
• Cross-Over Prediction Capability: Predicts PBRs in both structure-annotated and disorder-annotated proteins, addressing limitations of predictors trained exclusively on structured or disordered datasets.
• Reduction of Cross-Predictions: Minimizes incorrect predictions to non-protein partners such as nucleic acids and small ligands through selective combination of SCRIBER and disoRDPbind outputs.
• Benchmarking and Validation: Development and validation were informed by a comparative study using a benchmark dataset with both structure- and disorder-derived annotations and a representative collection of disorder- and structure-trained predictors.

Scientific Applications:

• Drug discovery: Enables identification of protein-binding residues relevant to mapping interaction sites for therapeutic targeting.
• Molecular biology: Supports functional studies of protein interactions by annotating binding residues in both structured and disordered regions.
• Bioinformatics: Provides sequence-based PBR annotations for comparative analyses and integrative computational studies of protein interactions.

Methodology:

Combines outputs from SCRIBER (trained on structured protein-protein complexes) and disoRDPbind (trained on disordered protein-binding residues) to balance and enhance PBR prediction accuracy across different protein types.