Improving phosphoproteomic kinase activity inference by contextual refinement of Kinase Library relationships

Protein-association filtering of Kinase Library–expanded kinase-substrate relationship priors improves phosphoproteomic kinase activity inference, retaining estimates for 308 kinases with AUROC > 0.75 and extending coverage to 200 additional kinases.

• Why it matters: Kinase activity inference depends on KSR priors, but Kinase Library expansion can reflect sequence compatibility rather than intracellular regulation, degrading discrimination, directionality, and ranking.
• What they did: Using 222 perturbation phosphoproteomic samples spanning 396 activation or inhibition events across 78 kinases, the study benchmarked KL-expanded priors and refined them by constraining KL-derived KSRs with predicted protein co-localization versus protein-association evidence.
• The result: Protein-association filtering consistently improved inference performance, reduced redundancy, and achieved a better accuracy–coverage balance than co-localization filtering, enabling structured, interpretable activity patterns and enrichment for dark kinases among newly scoreable targets.

AI summary of the abstract · may contain errors · not medical advice · read the paper.