Review papers with raw data transparency

What the paper did

• Generated a large dose‑resolved phosphoproteomics resource: 133 clinical kinase inhibitors × 5 human cancer cell lines × 11 doses → 665 decryptM experiments and ~17.1 million peptidoform dose‑response curves, profiling ~90,776 peptidoforms (Dataset scale and design).
• Developed potency coherence analysis (identity correlation icorr) to find potency-coherent KSRs and combined potency coherence with motif plausibility (Kinase Library) to nominate KSRs (hypothesis → candidate → confident) yielding 5,318 confident KSRs across 44 kinase groups and 1,809 proteins (Potency coherence and KSR discovery).
• Curated 836 seed peptidoforms for 44 kinase groups (covering 96 kinases) to calculate inf_pEC50 per experiment and thereby infer kinase activity changes and link to phenotype and patient phosphoproteomes (Seed peptidoform curation).

Major strengths

1. Sheer scale and public integration — 17.1M dose responses and ProteomicsDB integration allow continuous re‑evaluation and falsification of KSRs (ProteomicsDB integration).
2. Potency dimension is mechanistically sensible — using pEC50/potency to parse polypharmacology is physically motivated (drug KD maps to potency) and practical in deconvolving on/off‑targets across doses (Potency dimension rationale).
3. Controls for false positives — icorr statistical framework, multiple-testing correction, motif plausibility filter, and requirement for cross‑experiment potency coherence reduces many common annotation errors (Statistical controls and motif filter).

Key limitations and blindspots (evidence‑weighted)

• Limited chemical coverage — 133 clinical inhibitors do not span the 518 human kinases; off‑target inhibition and missing direct inhibitors create indistinguishability for kinases in the same pathway (authors note ERK vs RSKs, p38 vs MAPKAPKs issues) (Limited inhibitor coverage).
• Cell model sampling bias — five cancer cell lines provide useful heterogeneity but cannot capture tissue-specific kinase contexts; many kinases active in patients may be inactive in the chosen lines (Cell line scope limitation).
• Time window biases dynamic site detection — 100 minute treatments emphasise rapidly regulated phosphosites and miss slow or structural phosphorylation (authors flag static sites and slow kinetics) (100 minute treatment bias).
• Seed curation required manual input — 836 curated seeds enable inf_pEC50 but introduce expert choice and potential circularity; understudied kinases lacking seeds remain poorly annotated (Seed curation caveat).
• Phosphatase and network dynamics assumptions — potency coherence assumes phosphatase activity is not rate-limiting and that potency is preserved across indirect cascade steps; exceptions (priming kinases, conditional or shared substrates) can penalize true KSRs (Assumptions about phosphatase and shared sites).

Technical critique and reproducibility

Overall methods are comprehensive and reported with modern proteomics best practices (TMT11, Orbitrap LC‑MS3, MaxQuant, Percolator). The authors provide code and an icorr package (GitHub and PyPI) and integrate results into ProteomicsDB, supporting reproducibility; nevertheless full reproducibility requires public raw files, exact parameter files, and curated seed lists (Methods and reproducibility statements).

Independent validation: the authors show multiple orthogonal checks (missed‑cleavage ground truth for icorr, motif plausibility via Kinase Library, Kinobeads drug binding corroboration) — good practice but not replacement for targeted biochemical validation of novel KSRs (e.g., in vitro kinase assays, mutational site perturbations, substrate trapping) (Validation approaches used).

Actionable recommendations to strengthen and extend the work

• Expand chemical probe set beyond clinical inhibitors (selective probes for understudied kinases) to reduce n::1 assignment ambiguity and to separate pathway neighbors.
• Increase cell model diversity (DepMap panel, primary cells, non-cancer contexts) and extend time courses (short and long windows) to capture slow or structural phosphorylation.
• Publish seed peptidoform lists and curated code notebooks to enable independent re‑use and re‑curation by the community (the approach relies on curated seeds).
• Prioritize biochemical follow‑up on a selected set of novel confident KSRs (e.g., 10 high-confidence but previously unannotated edges) using orthogonal assays: recombinant kinase assays, substrate trapping, and phospho‑site mutagenesis in cells.
• Estimate effect of phosphatase activity explicitly (co‑treat with phosphatase inhibitors and repeat potency coherence) to test the assumption that phosphatase activity is not limiting.

Key scientific insights and likely impacts

The potency coherence framework reframes KSR discovery as a falsifiable, dose‑resolved statistical inference problem and provides the first large, dose‑resolved, proteome‑scale KSR map suitable for dynamic re‑evaluation. Practically, the 5,318 confident KSRs and the ProteomicsDB implementation will improve kinase activity inference in clinical phosphoproteomes and pharmacodynamic biomarker development — but success depends on continued expansion of chemical and cellular perturbation spaces (Potency coherence as evolving framework).

Reproducible visuals you can use immediately

Note The percent bins reflect the paper's reported inf_pEC50 summary: 9% regulated, 74% not regulated, 17% inconclusive (inf_pEC50 distribution).

Bioinformatics follow‑ups you can run (one‑click agent)

If you want full, reproducible reanalysis (seed list extraction, icorr re-run, alternate motif thresholds, cell-line stratified KSR lists, and patient kinase activity recalculation), run the BGPT AI biology agent below — it will execute pipelines (python, icorr, MaxQuant derived tables) and produce downloadable tables and figures.

Bottom line and confidence

Bottom line: This is a high‑quality, large‑scale, well‑documented resource that advances KSR discovery via a principled potency coherence framework and practical ProteomicsDB deployment; its value will grow as the chemical probe and cellular model space expand and as targeted biochemical follow‑up validates novel edges (Authors conclusion and outlook).

Author review links

Open bespoke author reviews on BGPT