Is MERTK on tumor cells in ZFTA-RELA but on macrophages in PFA?
MERTK dual-compartment
Does MERTK appear as a driver kinase differently across ZFTA-RELA (n=117) and PFA (n=123), and does that correlate with inflammatory pathway activity (myeloid infiltration proxy)?
✗ MERTK signal absent at this resolution
MERTK does not appear as a primary driver kinase in our ranker's top-10 across any EPN cohort. The compartmental hypothesis cannot be tested with pathway-level data; gene-level RNA + scRNA deconvolution would be required to evaluate it properly.
Per-cohort signal
| Cohort | n | MERTK as driver (top-10) | Inflammatory score median | corr(MERTK count, inflam) |
|---|---|---|---|---|
| ZFTA-RELA | 117 | 0.0% | 0.030 | — |
| PFA (broad) | 123 | 0.0% | -0.040 | — |
| Other EPN | 130 | 0.0% | — | — |
Why anti-PD-1 has failed in PFA — and what dual-compartment MERTK would change
Single-agent anti-PD-1 immunotherapy has repeatedly failed in posterior-fossa group A (PFA) ependymoma. The mechanism is well-documented: PFA tumors are immunologically cold — low tumor-infiltrating lymphocyte (TIL) density combined with an M2-skewed tumor-associated macrophage (TAM) compartment that actively excludes cytotoxic T cells. Checkpoint blockade has nothing to release.
The dual-compartment MERTK hypothesis flips this. MERTK is the dominant efferocytosis kinase keeping TAMs in an M2 / immunosuppressive state. A MERTK inhibitor would repolarize M2 → M1, restore T-cell infiltration, and prime the tumor for combination anti-PD-1 — converting a cold tumor into a hot one. Same drug class, different mechanism than ZFTA-RELA, where MERTK (if present on tumor cells) would act as a direct cytotoxic target.
Predicted treatment strategy by subgroup × compartment
Conditional on validating MERTK signal with gene-level RNA + compartment deconvolution (xCell/CIBERSORTx) or scRNA reference. Hypothetical map; not a treatment recommendation.
Per-sample distribution
MERTK-driver count = number of top-10 drugs (per sample) whose primary driver kinases include MERTK. Most cells are 0 here — see "Caveats" below.
| Sample | Cohort | Inflam score | MERTK driver count |
|---|---|---|---|
| No EPN sample has MERTK as a driver in any of its top-10 drugs. | |||
How this test runs (caveats)
PRD §EPN-4 prefers bulk-RNA deconvolution into tumor vs myeloid compartments (xCell/CIBERSORTx). That isn't in the precompute pipeline. We approximate compartmental MERTK by checking primary_driver_kinases from F-40's top-10 drug rankings, and use HALLMARK_INFLAMMATORY_RESPONSE as a myeloid-infiltration proxy. Treat as exploratory — a strong signal here justifies Frontiers 2022 scRNA ingestion, a null signal does not refute the hypothesis.
A null signal here does not refute the dual-compartment hypothesis — it tells us Pathway-level scoring (which compresses gene-level expression through GSVA pathway scores before ranking drugs) is the wrong lens. To evaluate this insight properly, KIRhub needs (a) per-sample gene-level RNA materialized into a queryable parquet, and (b) compartment deconvolution via xCell/CIBERSORTx or scRNA ingestion (Frontiers 2022, Gillen 2020, Gojo 2020).
# 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: mertk-dual-compartment