Research Use Only. KIRhub outputs are computational research artifacts. They are not validated for clinical decision-making, diagnosis, or treatment.
KIRhub 2.0· v2
Ependymoma Hub
V1V2
Sign in
KIRhub · Tumor hub

Pediatric ependymoma

Pick your role — tools and shortcuts tailored to your work.

+ Investigate an EPN sample

A pediatric brain tumor (~200 new US cases/year) whose biology splits cleanly into four molecular subgroups — PFA, PF-B, ZFTA-RELA, ST-YAP1 — with profoundly different prognoses and biology, despite identical histology.

🧪 Kinase & drug catalog — ependymoma

Compare selectivity across the kinome, drill into the pathway graph, or take ependymoma scope into the cohort scorer and atlas tools. Sample browser, ask the atlas, and strategic map are pre-filtered to EPN.

Browse kinases →Browse drugs →Pathway-reversal aggregate ↗Strategic map →Atlas tools ↗Drug calibration ↗

Subgroup landscape

ReferenceMeasuredModeled

PFA

Active research direction
Posterior Fossa A — H3K27me3-loss-driven; most aggressive pediatric subgroup.
Atlas n: ~152 samples
Age: Median age ~3–5 y · mostly pediatric
Prognosis: 5-y OS ~60%
SOC: Surgery + focal radiation. Chemo of limited utility. Recurrence common.
Lead hypothesis: CDK4/6 inhibitors (Palbociclib, Abemaciclib); hypoxia–wound niche multi-RTK.
Explore this subgroup →

ZFTA-RELA

Inconclusive — needs better data
Most common supratentorial subgroup; defined by the ZFTA-RELA gene fusion.
Atlas n: ~117 samples
Age: Mostly older children
Prognosis: 5-y OS ~70–80%
SOC: Surgery + radiation. No targeted therapy approved.
Lead hypothesis: Multi-RTK polypharmacology; MERTK on tumor cells.
Explore this subgroup →

PF-B

Not under investigation
Posterior Fossa B — chromosomal instability without H3K27me3 loss.
Atlas n: ~4 samples
Age: Older children / young adults
Prognosis: 5-y OS ~95%
SOC: Surgery + radiation. Favorable prognosis.
Lead hypothesis: None under active test — favorable prognosis lowers research priority.
Explore this subgroup →

ST-YAP1

Not under investigation
Supratentorial; driven by YAP1 fusions (YAP1-MAMLD1, YAP1-FAM118B).
Atlas n: ~10 samples
Age: Pediatric
Prognosis: 5-y OS ~100%
SOC: Surgery + radiation. Excellent prognosis.
Lead hypothesis: None under active test — outcomes already excellent.
Explore this subgroup →

What we tested, what we found

2 supported · 1 refuted · 2 inconclusive
ModeledSpeculative
PFA · Hypoxia–wound niche couplingRECOMMENDED NEXT
Supported

Hypoxic-core / wound-rim niche in PFA

The question
PFA tumors are theorized to have a hypoxic core surrounded by a wound-healing rim. Does that pattern actually show up in the data we have?
What we found
Yes — strongly. In PFA samples, hypoxia and wound-healing pathway scores correlate at r = 0.88, compared to r = 0.72 in non-PFA ependymoma.
Why it matters
If hypoxia and wound-healing are spatially coupled in PFA tumors, a multi-target drug that hits both could collapse the tumor's coupled niche.
⚠ Caveat
Bulk correlation cannot prove SPATIAL coupling. The decisive test is Visium spatial transcriptomics. Cabozantinib (the originally hypothesized drug) ranks poorly (68/92) for PFA.
→ Next step
Ingest Donson 2022 Visium. Parser is already written and unit-tested.
PFA · CDK4/6 inhibitor activity
Supported

CDK4/6 inhibitors in PFA

The question
Do CDK4/6 inhibitor drugs show up as good candidates for PFA samples in our drug-ranking engine?
What we found
Yes for Palbociclib (top-5 in 56% of PFA samples; median rank 2 of 92) and Abemaciclib (top-5 in 37%).
Why it matters
Abemaciclib is CNS-penetrant and has prior pediatric brain-tumor trial history (PNOC). Fastest path from compute to clinic.
⚠ Caveat
The mechanism story (developmental origin) was refuted. The drug-ranking finding stands as an empirical observation; the textbook explanation does not.
→ Next step
Retrospective re-analysis of PNOC abemaciclib trial data, stratified by pathway-reversal rank.
Developmental-origin framework test
Refuted

Developmental-origin framework falsified

The question
The field assumes pediatric ependymoma 'recapitulates fetal brain development'. Does the data support this?
What we found
No. With a proper adult-brain reference (GTEX, 1,409 samples), both pediatric and adult ependymoma cluster closer to adult brain than to fetal brain in hallmark pathway space.
Why it matters
This is the textbook mechanism story for several drug hypotheses. Refuting it doesn't kill the drug findings — but removes their mechanism story.
⚠ Caveat
Measured at HALLMARK pathway level. A test using neural-development-specific gene signatures could give a different answer.
→ Next step
Re-run the test against KEGG/Reactome neural-development pathways.
ZFTA-RELA · Multi-RTK polypharmacology
Inconclusive

Multi-RTK polypharmacology in ZFTA-RELA

The question
Do multi-target receptor kinase inhibitors (Cabozantinib, Tepotinib) concentrate as the ranker's top picks for ZFTA-RELA?
What we found
No. None of the 5 hypothesized drugs pass the subgroup-specificity gate. Top picks were CDK4/6, CHK1, JAK/BTK — same families as PFA.
Why it matters
If true, would be a clean rationale for a ZFTA-RELA-targeted multi-RTK trial. The test says we can't see it.
⚠ Caveat
Pathway-level ranker compresses gene-level RTK co-expression signal away. A null result doesn't refute biology — just says this lens can't resolve it.
→ Next step
Raw RNA per sample materialized into the ETL pipeline. Then re-test with gene-level RTK z-scores.
MERTK · Tumor vs macrophage compartment
Inconclusive

MERTK on tumor vs macrophages

The question
Is MERTK on tumor cells in ZFTA-RELA but on macrophages in PFA — a 'same drug, different mechanism' story?
What we found
Can't tell. MERTK never appears as a primary driver kinase in the ranker's top-10 for any of the 370 EPN samples (0%).
Why it matters
If validated, MERTK inhibitors would have a dual indication. The most clinically novel hypothesis on the table.
⚠ Caveat
Bulk pathway-level scoring cannot resolve tumor-vs-myeloid compartment. xCell/CIBERSORTx deconvolution or scRNA reference is the right tool.
→ Next step
Raw RNA + scRNA reference ingested.
Pediatric ependymoma · atlas at a glance
370
EPN samples in atlas
across 5 source cohorts
266
Pediatric (<18 y)
of 370
4
Molecular subgroups
PFA · PF-B · ZFTA-RELA · ST-YAP1
2/1/2
Hypothesis verdicts
supported · refuted · inconclusive
—
Catalog calibrated %
across full v2 drug catalog
Atlas: 1,358 brain-tumor samples (370 ependymoma, 888 medulloblastoma, 100 fetal-brain references) · 92 drugs ranked · 50 hallmark pathways scored.
Research Use Only · all insights are computational; none are validated for clinical decisions.