Predikin

Predikin predicts substrates of protein kinases, with emphasis on serine-threonine kinases, by analyzing kinase catalytic domains and substrate-determining residues to score potential phosphorylation sites.

Key Features:

• PredikinDB: A specialized database linking phosphorylation sites to specific protein kinase sequences to associate substrates with corresponding kinases.
• Perl module core: The analytical engine is implemented as a Perl module that performs the tool's core computations.
• Catalytic domain detection and classification: Locates protein kinase catalytic domains within sequences and classifies them by type or family.
• Substrate-determining residue identification: Identifies conserved residues within the kinase catalytic domain that contact the substrate near the phosphorylation site.
• Weighted scoring matrices: Generates weighted scoring matrices using three distinct methods to evaluate potential phosphorylation sites.
• Site extraction and scoring: Extracts putative phosphorylation sites from query sequences and scores these sites for specific kinases, with optional filters to refine predictions.
• SQL-enhanced prediction queries: Uses SQL queries on PredikinDB to generate and refine kinase–substrate predictions.
• Performance evaluation: Predictive performance has been evaluated using receiver operator characteristic (ROC) graph analysis.

Scientific Applications:

• Signal transduction analysis: Prediction of kinase substrates to investigate signaling pathways regulated by phosphorylation.
• Hypothesis generation for signaling mechanisms: Suggests candidate substrates to propose new kinase-mediated signaling interactions.
• Therapeutic target exploration: Identifies potential phosphorylation-mediated regulatory sites relevant to drug discovery.
• Cellular process characterization: Aids in understanding cellular processes controlled by serine–threonine phosphorylation.

Methodology:

Implemented as a Perl module that locates and classifies kinase catalytic domains, identifies substrate-determining residues contacting the substrate near phosphorylation sites, generates weighted scoring matrices by three methods, extracts and scores putative phosphorylation sites for specific kinases with optional filters, leverages SQL queries on PredikinDB for prediction generation, and has been evaluated using ROC graph analysis.