Comprehensive evaluation of phosphoproteomic-based kinase activity inference

Area of Science:

• Molecular Biology
• Bioinformatics
• Systems Biology

Background:

• Kinases are crucial regulators of cellular processes, implicated in numerous diseases.
• Inferring kinase activity from phosphoproteomics data is vital but limited by current methods and kinase-substrate libraries.
• Existing inference methods yield variable results, requiring robust evaluation for reliable interpretation.

Purpose of the Study:

• To introduce benchmarKIN, an R package for comprehensive evaluation of kinase activity inference methods.
• To benchmark kinase-substrate libraries, inference algorithms, and the utility of predicted interactions.
• To demonstrate the application of kinase activity inference in characterizing kinase inhibitor responses.

Main Methods:

• Development of benchmarKIN, an R package for evaluating kinase activity inference.
• Utilized classical perturbation experiments and a novel tumor-based benchmarking approach with multi-omics data.
• Evaluated kinase-substrate libraries, inference algorithms, and incorporated predicted kinase-substrate interactions (NetworKIN).

Main Results:

• Most computational inference methods performed similarly; however, library choice significantly impacted inferred kinase activities.
• Manually curated kinase-substrate libraries demonstrated superior performance in recapitulating known kinase activities.
• Integrating predicted targets from NetworKIN further enhanced performance in tumor-based evaluations.

Conclusions:

• benchmarKIN provides a robust framework for evaluating kinase activity inference methods.
• The selection of high-quality, manually curated kinase-substrate libraries is critical for accurate kinase activity inference.
• benchmarKIN aids researchers in selecting reliable methods for identifying deregulated kinases and understanding kinase inhibitor responses.