Skip to content

KRSA

Kinome Random Sampling Analyzer — end-to-end upstream-kinase analysis from PamChip kinome-array data — an R package and R Shiny application from the Cognitive Disorders Research Lab (CogDisResLab, University of Toledo) that automates peptide filtering, random-sampling permutation analysis, heatmap generation, and kinase-network visualization for PamChip functional-proteomics datasets.

Talk at GBCC2025

  • Speaker: Ali Sajid Imami (University of Toledo)
  • Day / session: Day 3 (Wed Jun 25, 2025) — lightning-talks block, 10:30–11:15 AM, Grace Auditorium (opening speaker)
  • Talk title: Demystifying kinome analysis through a streamlined R package designed for functional proteomics
  • Slides: TBD
  • Video: GBCC2025 playlist
  • Abstract / schedule: GBCC2025 program · Galaxy June 2025 meeting report

What it does

KRSA targets the specific data product of the PamGene PamChip system — a kinome microarray that measures peptide-substrate phosphorylation across hundreds of peptides as a proxy for kinase activity. The package automates the full upstream-kinase-analysis workflow: filtering noisy peptides, running random-sampling permutation tests to identify enriched upstream kinases, generating diagnostic heatmaps, and emitting kinase-network visualizations. The Shiny app makes the same pipeline accessible to wet-lab users who don't want to script R directly. It's a tightly scoped tool — built around one specific functional-proteomics modality — but exhaustive within that scope.

How it works

Core idea. For each candidate upstream kinase, KRSA repeatedly draws random subsets of peptides (matched in size to that kinase's known substrate set) from the full PamChip peptide universe and builds a null distribution of phosphorylation summary statistics, then scores the observed substrate set against that random background.

Inputs / outputs. Inputs are (1) raw PamChip kinome-array signal — median peptide-spot intensities across conditions — and (2) a kinase–peptide association file (e.g. from GPS 3.0 or Kinexus Phosphonet) that maps each peptide to its known upstream kinases. Output is a ranked kinase table with mean phosphorylation, SD, and Z-score against the random background, plus peptide heatmaps and a kinase-signaling-network visualization.

Key innovation. Random-sampling permutation as the null model: instead of assuming a parametric distribution of peptide-level effects, KRSA bootstraps a background per-kinase by sampling matched-size peptide sets, sidestepping the small-N problem typical of PamChip experiments and providing a Z-score with a directly empirical interpretation.

Parameters / API surface worth knowing. - Number of random-sampling iterations (controls null-distribution resolution). - Peptide filter thresholds (signal/quality cutoffs before analysis). - Kinase–peptide mapping database (GPS, Kinexus, or user-supplied). - Comparison contrast (which experimental conditions to score against each other).

Canonical example. From the project documentation: load PamChip median-intensity output → filter low-quality peptides → fit a per-peptide effect across conditions → run KRSA's random-sampling routine against a kinase-peptide table → emit a ranked kinase table with Z-scores plus a network plot showing the implicated signaling cascade. The Shiny app exposes the same pipeline through point-and-click.

Where it fits in the corpus

  • AACR 2026: axis = bioinfo / AI methods; functional-kinome readouts are upstream of any kinase-inhibitor pharmacology dossier
  • EuroBioC 2025: no direct counterpart on kinome arrays
  • ISMB 2026: scaffold
  • Nextflow Summit 2026: not mentioned

Notes

Imami opened the GBCC2025 lightning-talk session. The Imami group also has drugfindR (iLINCS drug-signature screening) under Bioconductor consideration, but the GBCC2025 lightning talk was specifically the kinome/PamChip tool — i.e. KRSA, not drugfindR.