KSP-PUEL

KSP-PUEL predicts novel substrates of specific kinases by integrating static kinase recognition motifs and dynamic mass spectrometry-based phosphoproteomics data within a positive-unlabeled ensemble learning framework.

Key Features:

• Integration of Static and Dynamic Data: Combines static kinase recognition motifs with dynamic phosphoproteomics measurements to capture both sequence-level and condition-specific phosphorylation information.
• Positive-Unlabeled Ensemble Learning: Implements a positive-unlabeled ensemble learning approach to improve sensitivity for novel substrate prediction while controlling specificity.
• Proteome-Wide Phosphorylation Quantification: Uses mass spectrometry-based proteome-wide phosphorylation quantification to inform kinase-specific substrate prediction.
• Application to Insulin Signaling Kinases: Applied to predict novel substrates of key kinases involved in insulin signaling pathways.

Scientific Applications:

• Signaling Network Reconstruction: Identifies kinase-substrate pairs to aid reconstruction of cellular signaling networks.
• Discovery of Novel Substrates: Predicts previously uncharacterized kinase substrates to accelerate experimental follow-up and hypothesis generation for targeted therapies.

Methodology:

KSP-PUEL extends positive-unlabeled learning into an ensemble model that integrates static kinase recognition motifs with dynamic phosphoproteomics data and was validated using simulation studies and real-world applications.