Oncofusion

start / graph / gene-ezhip

Genestrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q123-the-architecture-from-other-parts-analogy

EZHIP (CXorf67, CATACOMB) - a protein that jams PRC2 by pretending to be an H3K27M oncohistone, and the 3' payload of MBTD1's other fusion

also called CXorf67 · CATACOMB · EZH inhibitory protein · C11orf67-like naming error - NOT an alias, see caveats

EZHIP carries a short stretch of protein that looks like a mutated histone, and it uses that disguise to plug the enzyme cells rely on to switch genes off. It matters here because it is the half that a rare uterine cancer's fusion gene attaches to MBTD1 - the same MBTD1 that forms a leukaemia fusion with ZMYND11.

The same thing, for a scientist

**EZHIP is UniProt Q86X51 (EZHIP_HUMAN, 503 aa), HGNC:33738 at Xp11.22, primary Ensembl id ENSG00000187690.** `CXorf67` is its **PREVIOUS OFFICIAL SYMBOL** and `CATACOMB` an alias, and that matters practically: the entire fusion literature uses CXorf67 while the mechanism literature has migrated to EZHIP, so `MBTD1[tiab] AND EZHIP[tiab]` returns **0** and `MBTD1[tiab] AND CXorf67[tiab]` returns 4, while `EZHIP[tiab]` alone returns 62 against `CXorf67[tiab]`'s 21 [E1][E5]. **Mechanism.** A short, highly conserved region in the C-terminal part of the protein contains a methionine whose flanking sequence resembles the H3K27M oncohistone. That peptide contacts the active site of EZH2, blocks PRC2 methyltransferase activity and de-represses PRC2 targets; a single substitution of the methionine abolishes the effect [E2][E3][E4]. UniProt's feature table marks **401-409 as sufficient for EZH2 interaction and 403-423 as necessary and sufficient for inhibition** of PRC2/EED-EZH1 and PRC2/EED-EZH2 activity [E1]. The consequence is not global abolition of H3K27me3 but **loss of broad H3K27me3 domains with retention at CpG islands**, and H3K27me3-mediated allosteric activation of PRC2 is what makes the inhibition potent [E3]. In germ cells the same protein is described as **constraining** rather than switching off PRC2 [E4], so the physiological role is a tuner and the oncogenic role is a poison. **Why this gene has an entry in a leukaemia graph.** It is the 3' partner of **MBTD1::EZHIP** in low-grade endometrial stromal sarcoma [E6], and MBTD1 is the 3' partner of ZMYND11::MBTD1 in AML. The polarity reversal between those two fusions is this beat's sharpest disanalogy, and it cannot be stated without knowing what the sarcoma fusion's payload does. `fusion-mbtd1-ezhip` and `analogy-mbtd1-occupies-both-fusion-roles-polarity-not-parts` carry the comparison. **Coverage verdict, pasted rather than paraphrased.** `python scripts/is-symbol-covered.py EZHIP` on 2026-09-23, run alone: ``` EZHIP no match under 6 candidate names ``` and on the retired symbol, also run alone: ``` CXorf67 (retired or alias; current HGNC symbol is EZHIP) CXorf67 no match under 6 candidate names [WARNING: CXorf67 is also a PREVIOUS symbol of something] ``` **Both verdicts had to be re-taken one symbol at a time, because the script is order-dependent and gave a FALSE 'NOT IN HGNC under any current, previous or alias symbol - treat any answer as unsafe' for CXorf67 in a multi-symbol run.** See `ana-is-symbol-covered-is-order-dependent-and-reports-a-false-unresolvable`. **What this entry deliberately does not claim.** EZHIP has no established role in any leukaemia, no reported interaction with KMT2A, menin or any SEC component, and no dependency signal - DepMap screens it in only **75** cell lines with a **positive** mean gene effect of +0.178 [E7]. It is here as a fusion partner of a gene this graph already holds, and nothing more.

What it looks like
EZHIP (CXorf67, CATACOMB) - a protein that jams PRC2 by pretending to be an H3K27M oncohistone, and the 3' payload of MBTD1's other fusion503 aa
EZH2-interaction region · residues 401–409PRC2PRC2-inhibition region (H3K27M-like peptide) · residues 403–4230200400

No three-dimensional model is available for this protein, so the map above is the whole picture. That is itself informative: AlphaFold does not model chains this long, and the parts that have been crystallised are held under the entries for the individual domains.

Go deeper
The oncohistone-mimic mechanism, and the two things it is not

What it is. PRC2 deposits H3K27me2/3 through the SET domain of EZH2 or EZH1. The H3K27M oncohistone of diffuse midline glioma inhibits PRC2 by occupying that active site with a substrate it cannot turn over. EZHIP encodes a peptide that does the same thing without being a histone - a trans-acting, genetically encoded version of the same lesion, which is why overexpression of EZHIP and the H3K27M mutation produce similar chromatin profiles 23. In posterior fossa group A ependymoma, EZHIP overexpression is the hallmark, and the tumours lack H3K27me3 2.

What it is not, first: it is not an off switch. Jain et al. show retention of H3K27me3 at CpG islands with loss of the broad domains, and identify the allosteric loop as the reason - PRC2 activated by existing H3K27me3 is more inhibitable, which explains why spreading fails while nucleation survives 3. A bulk Western blot cannot see that distinction, and the only chromatin evidence for the MBTD1 fusion is a bulk Western blot 6. Any experiment on the fusion has to be designed for a redistribution.

What it is not, second: it is not only a cancer gene. Ragazzini et al. describe EZHIP constraining PRC2 activity in germ cells, and UniProt records inhibition of H3K27me3 deposition during spermatogenesis and oogenesis 14. So the protein has a physiological setting in which limited PRC2 inhibition is the point. A fusion that relocates a tuner is a different lesion from one that delivers an inhibitor, and the sarcoma literature has never distinguished them - which is one reason the fusion's mechanism is open rather than settled.

And a mechanism ambiguity carried inside UniProt's own record. The FUNCTION comment says EZHIP 'probably inhibits methyltransferase activity by limiting the stimulatory effect of cofactors such as AEBP2 and JARID2', citing the ependymoma paper 1. That is not the same mechanism as direct SET-domain occupancy, and the record carries both without adjudicating. Recorded as unresolved rather than collapsed into one story.

Where EZHIP is found, and the diagnostic apparatus that already exists for it

Tumour types named in the reviewed literature: posterior fossa group A ependymoma (where overexpression is the defining event), diffuse midline glioma, endometrial stromal sarcoma (as the MBTD1 fusion), non-small cell lung cancer and Merkel cell carcinoma 5. In 2026 EZHIP and H3K27me3 immunohistochemistry were also shown to mark CNS germinoma: across 126 germ cell tumour cases, EZHIP was positive in 95.1% (97/102) of germinomas and germinoma components with loss of H3K27me3 in 98.0% (100/102), against 0/24 EZHIP positivity in non-germinomatous components 8.

Why that last paper matters more than it looks. It establishes that both antibodies are routine, scoreable and validated on formalin-fixed paraffin-embedded tissue, which converts the single most informative open question about MBTD1::EZHIP - does a fusion-positive sarcoma lose H3K27me3? - from a project into a stain. exp-h3k27me3-and-ezhip-ihc-in-archival-mbtd1-ezhip-sarcoma states it.

One reagent caveat that is decisive and is not in any of these papers. Under this project's rule 74 an immunohistochemical readout is identified by its clone, not its label, and no clone was retrievable at abstract level for any EZHIP antibody in this beat. For a fusion retaining only the C-terminal half of EZHIP, an antibody raised against the N-terminus will not detect the fusion protein while correctly detecting overexpressed full-length EZHIP in every other tumour type. So an EZHIP-negative result on a fusion-positive sarcoma would be uninterpretable without epitope information, and the H3K27me3 arm is the one that can be read without it.

Facts and figures
uniprotQ86X51
length AA503
gene primaryEZHIP
hgncHGNC:33738
hgnc nameEZH inhibitory protein
ensemblENSG00000187690
cytobandXp11.22
accession resolutionHGNC fetched live 2026-09-23 returned a single record for EZHIP carrying Q86X51 and ENSG00000187690; UniProt Q86X51 returned EZHIP as the primary gene name with no gene synonyms. CXorf67 was confirmed as a PREVIOUS OFFICIAL SYMBOL by a HGNC prev_symbol search, which returned HGNC:33738 alone.
disorderUniProt annotates residues 1-72, 97-462 and 483-503 as disordered, i.e. essentially the whole protein outside the inhibitory region. There is no PDB entry for EZHIP in this entry's data and none was searched for.
relationship to this projectNONE DIRECT. EZHIP has no reported role in any leukaemia, no reported interaction with KMT2A, menin, LEDGF, DOT1L or any SEC component, and no fusion with any gene this graph held before cycle 83 other than MBTD1. It is written because it is the 3' payload of MBTD1's second fusion, and the MBTD1 polarity reversal is this beat's central disanalogy.
epigenetic readoutRemoves rather than writes. Inhibits EZH1/EZH2 methyltransferase activity through an H3K27M-mimicking peptide at 403-423, down-regulating H3K27me2 and H3K27me3. The expected spatial signature is loss of broad H3K27me3 domains with retention at CpG islands, potentiated by H3K27me3-dependent allosteric activation of PRC2.
cell typeGerm cells physiologically (spermatogenesis and oogenesis, by similarity in UniProt); ependymal-lineage tumour cells, germinoma cells and endometrial stromal sarcoma cells pathologically. HEK293T and HCT-116 for the fusion biochemistry. NO HAEMATOPOIETIC EXPRESSION OR FUNCTION IS REPORTED ANYWHERE IN THIS LITERATURE.
cell stateNOT CHARACTERISED in any setting relevant to this graph. In ependymoma the associated state is a PRC2-deficient, H3K27me3-low chromatin state with de-repression of neurodevelopmental genes; no self-renewal or differentiation-block phenotype is attributed to EZHIP itself.
subcellular localizationNOT MEASURED in this beat and not asserted. The protein acts on nuclear PRC2 and is inferred nuclear; no imaging or fractionation result was retrieved.
subnuclear localizationNOT MEASURED.
absence classliterature-absence-bounded
absence class noteThe absence this entry records is EZHIP in haematopoietic biology, and it is bounded by literature query alone: no archive was checked. The bounded counts are EZHIP[tiab] 62 and CXorf67[tiab] 21 records total across all diseases, with none in leukaemia among the titles inspected, and MBTD1[tiab] AND KMT2A[tiab] 0. NO GEO, PRIDE OR ProteomeXchange QUERY WAS RUN for EZHIP expression in haematopoietic cells, which is the check that would upgrade this to measured-not-analysed or downgrade it to never-measured. Recorded as literature-absence-bounded precisely so that a later cycle can see the archive half was not done rather than assume it was.
prev symbol
[
  "CXorf67"
]
alias symbol
[
  "CATACOMB"
]
domains
[
  {
    "name": "EZH2-interaction region",
    "start": 401,
    "end": 409,
    "pfam": null
  },
  {
    "name": "PRC2-inhibition region (H3K27M-like peptide)",
    "start": 403,
    "end": 423,
    "pfam": null
  }
]
essentiality
{
  "depmap_mean_gene_effect": 0.178,
  "depmap_n_screens": 75,
  "depmap_frac_below_minus_0_5": 0,
  "open_targets_dataversion": "26.06",
  "fetched": "2026-09-23",
  "interpretation": "NOT INTERPRETABLE AS A DEPENDENCY STATEMENT. 75 screens against 1,258 for MBTD1 and EPC1 in the same call, a positive mean, and a gain-of-function oncogenic mechanism that a knockout screen cannot address."
}
tumour types named in the literature
[
  "posterior fossa group A ependymoma - overexpression is the defining event, H3K27me3 lost",
  "diffuse midline glioma - alongside or instead of H3K27M",
  "low-grade endometrial stromal sarcoma - as the 3' partner of MBTD1::EZHIP",
  "CNS germinoma - EZHIP positive in 95.1% with H3K27me3 loss in 98.0% of 102 cases",
  "non-small cell lung cancer and Merkel cell carcinoma - named in a 2026 review, no primary source fetched here"
]
How this could bite us back
  • The obvious drug for a PRC2 protein is an EZH2 inhibitor, and for this gene it is backwards. Where EZHIP is the driver the lesion is already loss of PRC2 catalysis, so tazemetostat-class agents push in the direction the tumour went. Any Polycomb-inhibitor proposal for an EZHIP-driven or EZHIP-fusion tumour has to explain why more inhibition helps.
  • The inverse - restoring H3K27 methylation - has no agent. There is no EZH2 activator, and the H3K27M glioma field has pursued this far harder than a three-case sarcoma fusion ever will, without an approved epigenetic therapy. Recording the direction is honest; proposing it as a plan would not be.
  • EZHIP's inhibitory element is a twenty-residue peptide inside an otherwise disordered 503-residue protein with no structure. That is the least tractable shape a target can have for a small molecule, and a degrader needs a handle the disorder does not provide.
  • If the PALB2 arm is real, then blocking EZHIP could have a homologous-recombination consequence in normal germ cells, where the protein has a physiological job. That is an essential-function toxicity risk in a tissue nobody would think to monitor, and it is currently a single sentence in a review nobody has checked.
Still open
  1. Is EZHIP expressed in any haematopoietic cell? Nothing in this literature says, and the question is cheap - one GEO or Human Protein Atlas query. It matters because if it is silent in blood, then the MBTD1 promoter driving an EZHIP coding half is a plausible complete mechanism for the sarcoma fusion, and the comparison to a leukaemia fusion becomes a comparison of a promoter swap with a chimeric protein.
  2. Where does the EZHIP protein actually sit? Every mechanism paper measures its effect on a mark or on an enzyme; none reported in this beat measures where the protein is. An H3K27M oncohistone acts from inside chromatin and a soluble trans inhibitor acts from solution, and the two predict different spatial patterns of residual H3K27me3.
  3. Which antibody clone detects the C-terminal half of EZHIP? This is the gating question for every proposed immunohistochemical experiment on the sarcoma fusion, it is answerable from vendor datasheets, and rule 74 says the clone is the identifier. No clone was retrievable at abstract level in this beat.
  4. Is the PALB2 interaction real? A 2026 review reports that CXorf67 interacts with PALB2 and affects the BRCA1-PALB2-BRCA2 complex. No primary source for it was fetched here. If it holds, EZHIP is a DNA-repair as well as a chromatin protein and the MBTD1 fusion joins two DNA-repair-adjacent proteins, since MBTD1's own documented function in the TIP60 complex is homologous-recombination repair - which would be a second, entirely separate architectural story.
  5. Does the inhibitory peptide's disordered context make it undruggable, or is the EZH2 pocket it occupies the tractable side? The whole protein outside 403-423 is annotated disordered and there is no structure. Targeting the peptide is a protein-protein interface problem; targeting the pocket means an EZH2 ligand that excludes the peptide without inhibiting the enzyme, which nobody has attempted.
Every source behind this entry (9)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    EZHIP is HGNC:33738, 'EZH inhibitory protein', at Xp11.22, with CXorf67 as its previous official symbol and CATACOMB as an alias, primary Ensembl id ENSG00000187690 and UniProt Q86X51 (EZHIP_HUMAN), 503 amino acids; the feature table marks residues 401-409 as sufficient for interaction with EZH2 and 403-423 as necessary and sufficient for inhibition of PRC2/EED-EZH1 and PRC2/EED-EZH2 complex activity, annotates the protein as largely disordered (1-72, 97-462, 483-503), and the FUNCTION comment states that it inhibits EZH1/EZH2 methyltransferase activity causing down-regulation of H3K27me3, probably by limiting the stimulatory effect of cofactors such as AEBP2 and JARID2, and that it inhibits H3K27me3 deposition during spermatogenesis and oogenesis.

    What this evidence cannot show

    Both records fetched live 2026-09-23 with explicit field lists. gene_exact resolution was not needed because HGNC returns a single record for the symbol. The regions at 401-409 and 403-423 are curator summaries of in vitro reconstitution, not measurements made here, and 'necessary and sufficient' is the feature description's own wording. THE DISORDER ANNOTATION COVERS ALMOST THE WHOLE PROTEIN, which is a material fact about druggability and about whether any structural inference is available. A TOOL DEFECT WAS FOUND WHILE ESTABLISHING THIS RECORD: scripts/is-symbol-covered.py reported 'CXorf67 NOT IN HGNC under any current, previous or alias symbol - treat any answer as unsafe' in a multi-symbol run, which is false; see ana-is-symbol-covered-is-order-dependent-and-reports-a-false-unresolvable. The alias list in this entry's aliases field carries a deliberate warning string rather than a real alias, because no C11orf67 relationship exists and the near-miss is worth blocking.

    Evidence that argues against this (1)
    • The FUNCTION comment's stated mechanism - limiting cofactor stimulation, hedged with 'probably' - is not the same as the direct SET-domain occupancy the primary papers report, and the record cites the ependymoma paper for the cofactor version. A curation snapshot that carries two mechanisms without adjudicating is a pointer and not a source, and this project has already recorded a case where a UniProt feature table omitted a domain a paper had characterised. Jain SU et al. 2019
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    A small, highly conserved peptide in the C-terminal region of CXorf67/EZHIP mimics the sequence of K27M-mutated histones and binds the SET domain of EZH2, blocking methyltransferase activity and inhibiting PRC2 function with de-repression of PRC2 target genes including neurodevelopmental genes; overexpression of the gene is a hallmark of posterior fossa group A ependymoma, a tumour characterised by lack of H3K27me3 and by poor prognosis in infants and young children.

    PMID 30923826DOI
    What this evidence cannot show

    Abstract-level in this beat. Ependymoma; no MBTD1, no fusion, no sarcoma and no haematopoietic cell appears in it. The peptide-modelling arm is computational and the validation is immunocytochemistry, Western blot and methyltransferase assays. The first author's surname carries an umlaut in the source and is transliterated here. Genuinely new to this graph per check-citation-reuse. Identifiers fetched from PubMed 2026-09-23.

    Evidence that argues against this (1)
    • Two 2019 papers published within months established this mechanism with different emphases, and the disagreement is not trivial: this one locates the inhibition in SET-domain binding while also proposing interference with AEBP2 and JARID2 cofactor stimulation, and Jain et al. make the allosteric-activation loop central. A mechanism claim supported by three near-simultaneous papers that do not fully agree is established as a phenomenon and unsettled as a mechanism. Piunti A et al. 2019
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    A conserved sequence in EZHIP is necessary and sufficient to inhibit PRC2 catalytic activity in vitro and in vivo, contacting the EZH2 active site in a manner similar to the H3K27M oncohistone; expression of H3K27M or of EZHIP produces similar chromatin profiles, namely loss of broad H3K27me3 domains with retention of H3K27me3 at CpG islands; and H3K27me3-mediated allosteric activation of PRC2 substantially increases the inhibitory potential of both, providing a mechanism for the observed loss of H3K27me3 spreading in tumours.

    PFA ependymoma-associated protein EZHIP inhibits PRC2 activity through a H3 K27M-like mechanism. · Jain SU, Do TJ, Lund PJ, Rashoff AQ, et al. · Nat Commun 2019
    PMID 31086175DOI
    What this evidence cannot show

    Abstract-level in this beat, and cited principally for the SPATIAL refinement - loss of broad domains with CpG-island retention - because that is the detail that makes a bulk Western blot on the MBTD1 fusion insufficient. In vitro reconstitution plus cellular expression; no fusion and no sarcoma. Genuinely new to this graph. Identifiers fetched from PubMed 2026-09-23.

    Evidence that argues against this (1)
    • The equivalence of EZHIP and H3K27M profiles is drawn from ectopic expression in cell models, and the two lesions differ in a way the comparison suppresses: H3K27M is incorporated into chromatin as a nucleosome component and acts locally from within, while EZHIP is a soluble trans factor. Similar bulk profiles can arise from different spatial mechanisms, and nothing here measures where EZHIP protein itself is. Ragazzini R et al. 2019
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    EZHIP constrains Polycomb Repressive Complex 2 activity in germ cells, establishing a physiological setting in which the protein limits rather than abolishes H3K27 methylation.

    EZHIP constrains Polycomb Repressive Complex 2 activity in germ cells. · Ragazzini R, Perez-Palacios R, Baymaz IH, Diop S, et al. · Nat Commun 2019
    PMID 31451685DOI
    What this evidence cannot show

    TITLE-LEVEL PLUS THE UNIPROT FUNCTION COMMENT ONLY; the abstract was not analysed in this beat and no experimental detail, species or effect size is asserted. The claim stated here is deliberately no stronger than the title plus UniProt's 'By similarity' annotation for spermatogenesis and oogenesis. Cited because the physiological-tuner reading changes how a fusion delivering this protein should be modelled, and because it is the counterweight to reading EZHIP as an off switch. Identifiers fetched from PubMed 2026-09-23.

    Evidence that argues against this (1)
    • A title is not a result, and the distinction between 'constrains' and 'inhibits' may be the authors' framing of the same biochemistry rather than a different claim. Nothing in this item establishes that the germ-cell setting is mechanistically different from the tumour setting; it may simply be the same inhibition at physiological dose. Treating it as a separate mode would be over-reading a word in a title. Hubner JM et al. 2019
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.review

    A 2026 review reports that CXorf67 interacts with the PRC2 core components EZH2 and SUZ12 and thereby influences H3K27me3, that it is overexpressed in posterior fossa ependymoma, diffuse midline glioma, endometrial stromal sarcoma, non-small cell lung cancer and Merkel cell carcinoma, that in ependymoma and diffuse midline glioma it mimics H3K27M and inhibits PRC2, that in endometrial stromal sarcoma it forms fusion genes with MBTD1, and that its interaction with PALB2 affects the BRCA1-PALB2-BRCA2 complex and DNA repair.

    PMID 42238311DOI
    What this evidence cannot show

    A REVIEW, full text read this session via PMC13226146, cited as a map of tumour types and as the only record returned by MBTD1[tiab] AND H3K27me3[tiab]. It contains a DOCUMENTED ERROR reproduced in fusion-mbtd1-ezhip E13: it states MBTD1 'is associated with PRC1', where MBTD1's established membership is NuA4/TIP60. The PALB2/DNA-repair arm is reported here as the review reports it and NO primary source for it was fetched in this beat, so it is the weakest statement in this entry and nothing is built on it. Confidence moderate because a review carries no primary observation.

    Evidence that argues against this (1)
    • This review is demonstrably wrong about the complex membership of the gene it pairs EZHIP with, which lowers the weight of everything else in it that this beat did not independently check - including the tumour-type list and the PALB2 interaction. A secondary source that errs on a checkable primary fact should be used only for pointers, and the PALB2 claim in particular is a pointer and not evidence. Jacquet K et al. 2016
  6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Both CATACOMB/CXORF67 and the MBTD1-CXORF67 fusion protein interact with PRC2 and strongly decrease the catalytic products H3K27me2 and H3K27me3 in chromatin extracts; the fusion protein contains the majority of the MBTD1 sequence, lacking only its final 39 residues, joined to the C-terminal half of CATACOMB; and the inhibitory function maps to a short conserved region whose single methionine substitution abolishes it.

    PMID 31281901DOI
    What this evidence cannot show

    Full text read this session via PMC6609211. Every fusion experiment is ectopic HA-FLAG expression in HEK293T and HCT-116; the authors' own figure legend calls it the 'putative' fusion protein. No endometrial stromal sarcoma cell or tissue appears in the paper. The readout is a bulk-chromatin Western blot, which cannot resolve the redistribution E3 describes. Genuinely new to this graph.

    Evidence that argues against this (1)
    • The paper shows full-length CATACOMB producing the same H3K27me2/3 decrease as the fusion, so nothing in it establishes that the MBTD1 half contributes anything. On this evidence the fusion could be functionally an EZHIP overexpression event driven off the MBTD1 promoter, which would make the polarity comparison this graph builds on it a comparison of a chimeric protein with a promoter swap. Excluding that requires a cage mutant or an occupancy experiment, and neither exists. Hubner JM et al. 2019
  7. E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    DepMap via Open Targets, Open Targets dataVersion 26.06, fetched 2026-09-23: EZHIP has a mean gene effect of +0.178 across only 75 screens, with no screen below -0.5 and none below -1.0, against MBTD1 at -0.353 over 1,258 screens and EPC1 at -0.336 over 1,258 screens; the gene-specific denominator of 75 rather than 1,258 is itself the informative number, because it means DepMap cannot see this gene.

    What this evidence cannot show

    Rule 83 applies twice over. The RELEASE is recorded because the endpoint returns it; note however that meta.dataVersion is the OPEN TARGETS platform release and not a DepMap release name such as 24Q2, so this is labelled 'Open Targets dataVersion 26.06' rather than 'DepMap release 26/06'. The DENOMINATOR is gene-specific and here it is the whole story: 75 screens against 1,258 for the two comparators in the same call. A POSITIVE mean gene effect in 75 screens of an X-linked gene expressed in almost no somatic tissue is not evidence of dispensability - it is evidence that the assay has nothing to measure. Nothing in this entry should be read as a dependency claim in either direction.

    Evidence that argues against this (1)
    • A CRISPR knockout screen is the wrong instrument for this gene in a second way as well: EZHIP's oncogenic action is a GAIN of function - overexpression of an inhibitory peptide - so even a complete screen at 1,258 lines would be asking whether cells need EZHIP rather than whether they are harmed by it. The informative screen would be overexpression, and DepMap does not run one. Jain SU et al. 2019
  8. E8strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    EZHIP and H3K27me3 immunohistochemistry are validated, routine assays on formalin-fixed paraffin-embedded tissue: across 126 central nervous system germ cell tumour cases, EZHIP was positive in 95.1% (97/102) of germinomas and germinoma components with loss of H3K27me3 in 98.0% (100/102), against 0/24 EZHIP positivity and 12.5% H3K27me3 loss in non-germinomatous components, giving EZHIP 95.1% sensitivity with 100% specificity and H3K27me3 loss 98.0% sensitivity with 87.5% specificity.

    PMID 42453183DOI
    What this evidence cannot show

    Abstract-level. Cited ONLY to establish that the two stains are routine, scoreable and validated on archival material. The disease is a CNS germ cell tumour and bears no relation to endometrial stromal sarcoma. Denominators differ between markers because not every case was stained for every marker (102, 82, 97). NO ANTIBODY CLONE IS RECORDED, because none was retrievable at abstract level - and under rule 74 that is the identifier that matters, so this item supports 'the assay exists and performs' and not 'this reagent will work on a fusion protein'. Genuinely new to this graph.

    Evidence that argues against this (1)
    • Performance on EZHIP OVEREXPRESSION says little about detection of an EZHIP FUSION retaining only the C-terminal half. If the antibody's epitope is N-terminal it will miss the fusion protein entirely while performing perfectly in every tumour it was validated on - an epitope-scoped false negative with no visible signature. Any use of this assay on a fusion-positive sarcoma must either establish the epitope or read H3K27me3 alone. Piunti A et al. 2019
  9. E9strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    Coverage and naming verdicts, 2026-09-23, each run as a single-symbol invocation: scripts/is-symbol-covered.py EZHIP returns 'EZHIP no match under 6 candidate names'; scripts/is-symbol-covered.py CXorf67 returns 'CXorf67 (retired or alias; current HGNC symbol is EZHIP)' followed by 'CXorf67 no match under 6 candidate names [WARNING: CXorf67 is also a PREVIOUS symbol of something]'; scripts/is-symbol-covered.py CATACOMB returns 'CATACOMB (retired or alias; current HGNC symbol is EZHIP)' followed by 'CATACOMB no match under 6 candidate names'. Bounded PubMed counts the same day: EZHIP[tiab] 62; CXorf67[tiab] 21; CATACOMB[tiab] 20 and contaminated by the ordinary English word; MBTD1[tiab] AND EZHIP[tiab] 0; MBTD1[tiab] AND CXorf67[tiab] 4; EZHIP[tiab] AND PRC2[tiab] 22; EZHIP[tiab] AND H3K27me3[tiab] 33.

    What this evidence cannot show

    EVERY is-symbol-covered VERDICT HERE WAS RE-TAKEN AS A SINGLE-SYMBOL INVOCATION, because the script is order-dependent: in a multi-symbol run the first symbol that reaches the no-match branch overwrites the retired/alias index with a string, after which a retired symbol is either mis-reported as 'NOT IN HGNC under any current, previous or alias symbol - treat any answer as unsafe' or raises AttributeError. My own first run produced exactly the false verdict for CXorf67. See ana-is-symbol-covered-is-order-dependent-and-reports-a-false-unresolvable. The PubMed side used one [tiab] tag per term, never the parenthesised group form; CATACOMB[tiab] is reported as contaminated and was not used as a bare count. Titles and abstracts only.

    Evidence that argues against this (1)
    • A no-match verdict is not a gap - the script's own closing line says so - and the thing may live inside a complex, mechanism or mark entry with no name of its own. For EZHIP that caveat has real force: PRC2 and H3K27me3 both have entries in this graph, and an EZHIP-shaped fact could already sit inside mark-h3k27me3 or a PRC2-arm entry without the symbol appearing. This beat did not read those entries in full before writing, so the novelty claim rests on the id-and-alias scan and not on a reading of the PRC2 arm. oncofusion-knowledge-graph tooling 2026
Connected to