Research Use Only. KIRhub outputs are computational research artifacts. They are not validated for clinical decision-making, diagnosis, or treatment.
KIRhub 2.0· v2Sign in
Overview·PFA radial-glia
supported

Do CDK4/6 inhibitors concentrate in PFA ependymoma?

ModeledSpeculative
Question
Posterior-fossa A (PFA) is the most common and deadliest pediatric EPN subtype. The theory: PFA tumors are stuck in a fetal-development state that depends on CDK6. Test: do CDK4/6-inhibitor drugs (Palbociclib, Abemaciclib, Ribociclib) actually rank highly for PFA samples?
What we found
Yes for two of them: Palbociclib lands in the top 5 best-fit drugs for 56% of PFA samples (median rank 2 out of 92). Abemaciclib lands top-5 in 37% (median rank 13). Ribociclib does not. The subgroup is real, even if modest in size.
Caveat
The mechanism story underneath this insight (PFA = stuck in fetal development → CDK6-dependent) was just refuted by the falsification test. The drug-ranking finding stands as an empirical fact; the textbook 'why' does not. Also, Palbociclib appears in non-PFA EPN at similar rates — the subgroup specificity is weaker than ideal.
Next step
Retrospective re-analysis of PNOC abemaciclib trial data, stratified by pathway-reversal rank. Does the high-rank subset show better response than the original trial saw across all comers?
Scroll for the data, methods, and per-sample detail.

PFA radial-glia CDK4/6 → Abemaciclib

Do CDK4/6 inhibitors concentrate in PFA samples (n=124, broad) relative to PF-B (n=4), PF_SE (n=7), and other EPN (n=235)?

✓ Hypothesis supported

At least one CDK4/6 inhibitor passes the PFA-specificity gate (top-5 in ≥40% of PFA with median rank ≤ 15). PNOC/COG abemaciclib retrospective re-analysis is justified.

CDK4/6 inhibitor head-to-head

ModeledSpeculative

Named drugs checked against the gate (top-5 in ≥40% of PFA AND median rank ≤ 15).

DrugPFA top-5PFA median rankOther-EPN top-5Other-EPN median rankTop drivers (PFA)Gate
Abemaciclibcalibrated (0%)37.1%
1330.6%56GSK3B,GSK3A,CAMK2A,PIM1,EGFR✗ fail
Palbociclibcalibrated (0%)56.5%
236.6%70DYRK1B,PLK1,GSK3B,CAMK2A,EIF2AK2✓ pass
Ribociclibcalibrated (0%)0.8%
351.3%49LYN,EGFR,GSK3A,CAMK2B,CAMK2D✗ fail

What the ranker picks for PFA — top 25 drugs by top-5 frequency

ModeledSpeculative
DrugPFA top-5PFA median rankOther-EPN top-5Primary driversCDK4/6?
Encorafenibcalibrated (0%)58.9%432.8%GSK3B,GSK3A,NLK,LYN,RAF1
Palbociclibcalibrated (0%)56.5%236.6%DYRK1B,PLK1,GSK3B,CAMK2A,EIF2AK2
Abemaciclibcalibrated (0%)37.1%1330.6%GSK3B,GSK3A,CAMK2A,PIM1,EGFR
Tofacitinibcalibrated (0%)26.6%6930.6%JAK2,JAK1,TYK2,LYN,EGFR
Zanubrutinibcalibrated (0%)24.2%8433.2%EGFR,LCK,LYN,FYN,CSK
Darovasertibcalibrated (0%)21.8%127.7%GSK3B,NLK,GSK3A,LYN,PIM1
Gedatolisibcalibrated (0%)18.5%146.8%GSK3B,EGFR,FYN,LCK,PRKX
Tucatinibcalibrated (0%)18.5%5918.3%JAK1,JAK2,LYN,GSK3B,LCK
Rabusertibcalibrated (0%)18.5%2926.4%CAMK2A,CAMK2D,GSK3B,LYN,RAF1
Temsirolimuscalibrated (0%)15.3%279.4%EGFR,GSK3B,GSK3A,ABL1,WEE1
Mobocertinibcalibrated (0%)14.5%5415.7%EGFR,LYN,ABL1,GSK3B,FYN
Defactinibcalibrated (0%)13.7%354.7%GSK3B,EGFR,LCK,JAK2,PLK1
Tepotinibcalibrated (0%)13.7%223.4%GSK3B,LYN,ABL1,LCK,AXL
Sirolimuscalibrated (0%)12.1%236.4%EGFR,GSK3B,GSK3A,ABL1,AKT1
Ibrutinibcalibrated (0%)12.1%8116.6%LYN,EGFR,FYN,LCK,CSK
Deucravacitinibcalibrated (0%)12.1%232.6%GSK3B,LYN,ABL1,LCK,FGFR1
Duvelisibcalibrated (0%)11.3%223.8%LYN,GSK3A,ABL1,EGFR,AKT1
Entrectinibcalibrated (0%)11.3%8217.4%LYN,FYN,LCK,ABL1,JAK2
Umbralisibcalibrated (0%)9.7%6411.5%LYN,LCK,JAK2,JAK1,ABL1
Futibatinibcalibrated (0%)8.9%518.9%JAK2,FGR,JAK1,FGFR1,RET
Netarsudilcalibrated (0%)8.9%2615.3%AKT1,PRKX,ROCK1,AKT3,LRRK2
Alpelisibcalibrated (0%)8.9%588.5%EGFR,JAK2,JAK1,ABL1,GSK3B
Quizartinibcalibrated (100%)8.1%3017.4%GSK3B,FYN,PDGFRB,RET,FLT3
Asciminibcalibrated (0%)7.3%147.2%GSK3B,ABL1,AKT1,LYN,JAK2
Everolimuscalibrated (0%)5.6%265.1%EGFR,GSK3A,ABL1,AKT1,WEE1

How this test runs (caveats)

Radial-glia signature (HES1/FABP7/SLC1A3/PAX3/ZIC1/MKI67/CDK6) from raw RNA is the PRD's preferred input. Raw RNA isn't in the precompute pipeline; this test uses F-40 selectivity-adjusted rank for CDK4/6 inhibitors against the full KIRhub panel as the subgroup-specificity proxy. CDK6/CDK9 in primary_driver_kinases is the gene-level signal we CAN detect.

PFA-broad cohort = EPN_classification = PF-A OR EPN_subtype = "Posterior Fossa EPN" without PF-B/PF_SE marker. The explicit PF-A count is 17; with the broad definition we get n=124. Further ingestion of Pajtler 2015 would expand explicit PF-A to ~200.

Research Use Only · dataset saifudeen2026 1.0.0 · api 0.1.0 · 11 ms
# Platform
Saifudeen, A., et al. (2026). KIRhub: a falsification-first research workbench for translational oncology. Nature Biotechnology. https://doi.org/10.1038/s41587-026-03090-8

# Data source
McFerrin, L. G., et al. (2018). Oncoscape: a tool for interactive cancer genomics data analysis. Nature Genetics.
Arora, S., et al. (2026). A pan-pediatric brain tumor reference map: medulloblastoma and ependymoma in shared UMAP space. Neuro-Oncology.

# Datasets used in this insight
Oncoscape pediatric brain tumor compendium (n=1,358)

# Provenance
insight_id: pfa-cdk46