Data sources, licenses, and citations

KIRhub is built on data from peer-reviewed publications and public databases. Every dataset surfaced in the platform is listed here with its citation and license. If you publish or present work based on KIRhub output, please cite the original sources below — not KIRhub alone.

Aggregate license posture

KIRhub is licensed Apache 2.0. The combined corpus inherits the most restrictive license of any included dataset: non-commercial research use is unambiguously permitted across all sources. Commercial deployment requires additional licensing (Reaction Biology for the inhibition data, KEGG for that pathway library, confirmation of Broad TOU for CCLE / PRISM). PhosphoSitePlus is deliberately excluded from KIRhub's production scoring chain per PRD §4.1.3 (citation bias) — see section 7 below for the audit.

The "Research Use Only" badge in the header reflects this stance.

What the data-characteristic labels mean

Throughout KIRhub, each insight, verdict, and score carries small labels describing the nature of the data behind it, alongside an “i” explainer that opens its full methodology (data sources, process, math, thresholds, and caveats). This is the legend for those labels.

Measured: Comes directly from a primary dataset or laboratory assay.
Derived: Computed deterministically from measured data (e.g. a matrix product, cosine, or median).
Modeled: An algorithmic prediction — an estimate produced by the model, not a direct observation.
Calibrated: The model output has been validated against an external benchmark (CCLE/PRISM cell-line response).
External: Fetched live from a third-party source outside KIRhub.
Reference: Drawn from a curated public knowledge base.
Speculative: No validation yet — treat as an exploratory hypothesis, not a finding.

1. Saifudeen 2026 — KIRhub original profiling

Provides: Drug × kinase inhibition profiles for 92 clinical kinase inhibitors against 384 WT kinases + 342 variants. Dose-response curves, selectivity scores. The bedrock of every F-40 score. Also Table S9 — a 92-drug × 47 mesenchymal-specific-kinase residual-activity matrix (1 µM) powering EMT / mesenchymal druggability (previously unused, wired in 2026-06).
Source: DOI: 10.1038/s41587-026-03090-8 · Nature Biotechnology (2026)
License: Inhibition assay data © Reaction Biology Corporation. Supplementary tables CC-BY 4.0 per Nature Biotechnology policy.
Required citation: Saifudeen Z, Zhu L, Liang J, et al. Comprehensive profiling of clinical kinase inhibitors against wild-type and variant kinases. Nature Biotechnology (2026). doi:10.1038/s41587-026-03090-8
Where it shows up: Drug catalog
Dose-response curves
F-40 reversal scoring (D matrix)
EMT / mesenchymal druggability (Table S9)
EMT druggability + reversal cross-validation (Table S9)

2. Oncoscape — patient tumor cohorts

Provides: GSVA pathway scores for ~9,200 patient samples across bladder, brain, breast, colon, head-neck, lung, melanoma, meningioma, medulloblastoma + ependymoma.
Source: Oncoscape public S3 bucket. Per-cohort manifest YAML records the original publication for each cohort.
License: Per-cohort. Most are publicly redistributable for non-commercial research. Defer to the per-cohort manifest for the canonical citation.
Required citation: Holland Lab medulloblastoma + ependymoma — Hovestadt V et al., Nature (2014) and follow-ups.
TCGA cohorts (Bladder, BRCA, COAD, HNSC, LUAD/LUSC, SKCM) — The Cancer Genome Atlas Research Network, multiple flagship papers.
Meningioma — Choudhury A et al., Nature Genetics (2022).
Where it shows up: Cohort browser
Pathology-free tumor map
Cohort scoring

3. CCLE — Cancer Cell Line Encyclopedia

Provides

RPKM expression, clinical annotations, drug IC50, mutations for 1,156 cell lines. Used for calibration (CCLE IC50) and for synthetic CCLE tumor cohorts (AML / CML / HCC).

Source

Original Broad Institute releases (CCLE 2012, CCLE 2019). KIRhub reads via the cBioPortal CCLE mirror.

License

Non-commercial research use. Broad Institute Terms of Use apply.

Required citation

Barretina J et al. The Cancer Cell Line Encyclopedia enables predictive modelling of anticancer drug sensitivity. Nature 483:603–607 (2012). doi:10.1038/nature11003

Ghandi M et al. Next-generation characterization of the Cancer Cell Line Encyclopedia. Nature 569:503–508 (2019). doi:10.1038/s41586-019-1186-3

Where it shows up

4. PRISM Repurposing — drug viability

Provides: Viability fold-change for 236 drugs × 1,065 cell lines (201,810 rows). Second calibration source alongside CCLE IC50.
Source: PRISM Repurposing Secondary Screen, Broad Institute.
License: Non-commercial research use (Broad TOU).
Required citation: Corsello SM et al. Discovering the anti-cancer potential of non-oncology drugs by systematic viability profiling. Nature Cancer 1:235–248 (2020). doi:10.1038/s43018-019-0018-6
Where it shows up: Calibration ribbon per_source = prism_viability

5. OmniPath — kinase-substrate edges

Provides

Curated enzyme-substrate edges. Kinase-phosphorylation subset = ~39,000 edges in Operator T.

Source

omnipathdb.org REST API

License

Academic Free License (AFL) — Apache 2.0 compatible. Commercial use permitted with attribution.

Required citation

Türei D, Korcsmáros T, Saez-Rodriguez J. OmniPath: guidelines and gateway for literature-curated signaling pathway resources. Nature Methods 13:966–967 (2016). doi:10.1038/nmeth.4077

Türei D et al. Integrated intra- and intercellular signaling knowledge for multicellular omics analysis. Molecular Systems Biology 17:e9923 (2021).

Where it shows up

6. NetworKIN — kinase substrate predictions

Provides: Machine-learning predictions of kinase-substrate pairs with confidence scores. Scaffolded in build_T; remote was unavailable at last build.
Source: networkin.info v3.0
License: Academic use. Commercial use requires permission.
Required citation: Linding R et al. NetworKIN: a resource for exploring cellular phosphorylation networks. Nucleic Acids Research 36(Database):D695–D699 (2008). doi:10.1093/nar/gkm902
Where it shows up: Operator T (when populated)

7. PhosphoSitePlus — DELIBERATELY EXCLUDED from production

Provides: Literature-curated phosphorylation sites with high citation density. KIRhub's production Operator T deliberately excludes PSP per PRD §4.1.3 to avoid citation bias.
Source: phosphosite.org — registration required
License: PhosphoSitePlus non-commercial license. KIRhub uses PSP only in audit mode (never in production T) so the NC license does not block downstream commercial use of KIRhub itself.
Required citation: Hornbeck PV et al. PhosphoSitePlus, 2014: mutations, PTMs and recalibrations. Nucleic Acids Research 43:D512–D520 (2015). doi:10.1093/nar/gku1267
Where it shows up: Audit-only operator T (see PSP_AUDIT_REPORT.md in repo)

8. MSigDB — Hallmark, Reactome, KEGG

Provides

Three pathway libraries for Operator P. Hallmark (50 pathways, default), Reactome (~1,500), KEGG (~190). KIRhub uses all three for the substrate × pathway projection.

Source

gsea-msigdb.org v2024.1.Hs

License

Hallmark + Reactome subsets are CC-BY 4.0. KEGG content retains KEGG's own licensing — non-commercial use only without a KEGG commercial license.

Required citation

Liberzon A et al. The Molecular Signatures Database (MSigDB) hallmark gene set collection. Cell Systems 1:417–425 (2015). doi:10.1016/j.cels.2015.12.004

Subramanian A et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. PNAS 102:15545–15550 (2005).

Reactome subset: Jassal B et al. The Reactome pathway knowledgebase. Nucleic Acids Research 48:D498–D503 (2020). doi:10.1093/nar/gkz1031

KEGG subset: Kanehisa M, Goto S. KEGG: Kyoto Encyclopedia of Genes and Genomes. Nucleic Acids Research 28:27–30 (2000). doi:10.1093/nar/28.1.27

Where it shows up

9. HGNC — HUGO Gene Nomenclature Committee

Provides: Canonical gene symbols. Used to normalize kinase / substrate names across all sources.
Source: genenames.org
License: CC0 (public domain).
Required citation: Seal RL et al. Genenames.org: the HGNC resources in 2023. Nucleic Acids Research 51:D1003–D1009 (2023). doi:10.1093/nar/gkac888
Where it shows up: Every kinase symbol shown in KIRhub

10. ClinicalTrials.gov

Provides: Trial metadata for drug / kinase / mutation searches.
Source: ClinicalTrials.gov public API
License: Public domain (U.S. federal). No restriction.
Required citation: U.S. National Library of Medicine. ClinicalTrials.gov.
Where it shows up: Live trial search

How to cite KIRhub itself

KIRhub is built on the Saifudeen 2026 dataset and on the public sources above. When citing a specific F-40 score or atlas finding in a publication, please cite (a) the original underlying dataset(s), and (b) the model release identifier shown on the score response (e.g. f40@2026-06-10) — this fully specifies the operator T release, pathway library, and selectivity weight used.

Every score response carries a reproducibility URL. Including that URL in your supplementary materials lets reviewers re-derive the exact ranking you reported.