KinPred

KinPred unifies and standardizes human kinase–substrate prediction datasets to enable proteome-level mapping and filtering of tyrosine and serine/threonine kinase–substrate relationships.

Key Features:

• Unified prediction framework: Standardizes and integrates predictions from multiple published kinase–substrate resources into a common representation.
• Reference proteome mapping: Consistently maps substrates and kinases to current human reference proteomes to ensure identifier concordance across datasets.
• Phosphoproteome filtering: Filters predictions for relevance to the human phosphoproteome to retain experimentally plausible phosphorylation sites.
• Coverage of kinase classes: Includes predictions for tyrosine kinases and serine/threonine kinases at the proteome level.
• Algorithmic feature retention: Preserves algorithm-specific bases such as sequence specificity of kinase catalytic domains, evolutionary relationships, coexpression patterns, and protein–protein interaction networks.
• Comparative analysis support: Enables exploration of study bias, emergent network properties, and comparative analyses within and between predictive algorithms.
• Update mechanisms: Provides standardized methods to incorporate future updates aligned with expanding human phosphoproteome data.

Scientific Applications:

• Phosphoproteomic interpretation: Prioritizes candidate regulatory kinases for dysregulated phosphorylation sites in human phosphoproteomic datasets.
• Kinase–substrate discovery: Facilitates identification of potential kinase assignments for novel phosphorylation sites across the human proteome.
• Algorithm comparison: Supports comparative benchmarking of prediction algorithms and investigation of algorithm-specific network properties.
• Bias and network analysis: Enables analysis of study biases and emergent properties in integrated kinase–substrate networks.

Methodology:

Map substrates and kinases from multiple predictive resources to a common current human reference proteome; standardize identifiers and prediction formats; filter predictions against the human phosphoproteome; integrate (unify) predictions from three existing resources; and provide procedures for future updates and comparative network analyses.