Oncofusion

start / graph / hyp-both-mbtd1-fusions-converge-on-an-h3k27me3-poor-state

Hypothesisspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.Q123-the-architecture-from-other-parts-analogy

Both MBTD1 fusions end at H3K27me3-poor chromatin, one by excluding PRC2 and one by poisoning it - so the shared endpoint is the phenotype and MBTD1 is the coincidence

also called the H3K27me3 convergence · exclusion versus antagonism

MBTD1 is part of two different cancer fusion genes, and both of them appear to end up with the same missing chemical tag on DNA - the tag that keeps genes switched off - but they get there in opposite ways. One builds a chemical environment the silencing machine cannot invade; the other carries a decoy that jams the machine directly.

The same thing, for a scientist

**Two observations, neither citing the other.** In AML, ZMYND11::MBTD1 binds 16,114 sites that overlap H3K36me3 and are **devoid of H3K27me3**, and the authors propose that the resulting H3K36me3-plus-acetylation state is **refractory to PRC2 invasion** [E1]. In low-grade endometrial stromal sarcoma, MBTD1::EZHIP delivers an H3K27M-mimicking peptide that binds the EZH2 active site, and the fusion protein **strongly decreases H3K27me2/3** [E2][E3]. **The hypothesis.** The H3K27me3-poor state is the functional endpoint both fusions are selected for, and the two mechanisms - passive exclusion by an incompatible active mark, active catalytic antagonism - are two routes to it. On this reading MBTD1's presence in both fusions is a coincidence of gene neighbourhood and breakpoint availability, not a mechanistic thread. **Why it is worth writing rather than just noticing.** It makes two predictions that cut in opposite directions and are both cheap. If the endpoint is what matters, then **H3K27me3 should be depleted in fusion-positive sarcoma tissue** in the pattern EZHIP produces elsewhere - loss of broad domains with retention at CpG islands [E4] - and that is one stain on archival material with a validated antibody [E5]. And **H3K36me3 and H3K27me3 should be anti-correlated at ZMYND11::MBTD1 sites in the deposited leukaemia data**, which is pure compute on public files. **Why it is probably wrong in its strong form, and the entry says so up front.** H3K36me3 and H3K27me3 are antagonistic throughout the genome, in every cell type, independently of any fusion - the SETD2/PRC2 antagonism is textbook. **So 'fusion sites are H3K36me3-high and H3K27me3-low' may be a restatement of where the fusion binds and carry no information about the fusion at all.** That is a base-rate problem of exactly the shape this project has documented four times, and the analysis below is designed around it: the comparator is not zero, it is **H3K36me3-matched non-fusion-bound regions**. If the depletion at fusion sites is no greater than at matched sites, the leukaemia arm of this hypothesis is a null dressed as an observation. **Prior probability 0.15**, and the low value is mostly the base-rate problem rather than doubt about either measurement.

Go deeper
The two arms, and the comparator that decides each

Arm 1, the sarcoma, and it is the informative one. No chromatin measurement of any kind exists in MBTD1::EZHIP tumour tissue - MBTD1[tiab] AND H3K27me3[tiab] returns exactly one record and it is a 2026 review 6. The fusion protein's effect on H3K27me2/3 is a bulk Western blot on an ectopic construct in HEK293T 2. The prediction is specific and the assay is routine: H3K27me3 immunohistochemistry on archival fusion-positive tumour, scored against fusion-negative low-grade ESS. The antibody is in diagnostic use and was reported in 2026 at 98.0% sensitivity for H3K27me3 loss across 102 CNS germ cell tumour cases 5.

And the comparator matters more than the case. Low-grade ESS is dominated by JAZF1-SUZ12, which fuses a NuA4 component to a PRC2 core subunit 3. Whether those tumours retain H3K27me3 is the question that makes the MBTD1::EZHIP result interpretable - if the whole family is H3K27me3-low, the mark is a property of the tumour type; if only the EZHIP-fusion cases are, it is a property of the lesion. A single stain across a low-grade ESS series answers both at once, and the series exists in any gynaecological pathology archive.

Arm 2, the leukaemia, and it is the one that will probably return a null. The claim is that ZMYND11::MBTD1's target state actively excludes PRC2. The published observation is only that fusion peaks lack H3K27me3 1. Three comparators are needed and none has been run: (i) H3K36me3-matched regions not bound by the fusion, which is the base-rate control; (ii) the same regions before fusion expression, which needs a time course nobody has done; (iii) H3K27me3 after fusion withdrawal, which is the only measurement that could show exclusion is active rather than incidental. Arm 2 is measured-not-analysed for comparator (i) and never-measured for the other two.

What a shared endpoint would and would not mean for this project

If it holds, the interesting consequence is not about MBTD1. It is that two unrelated fusion mechanisms - reader-to-acetyltransferase and scaffold-to-PRC2-antagonist - converge on removing the same repressive mark, in two different tissues. That would make H3K27me3 depletion a fusion-oncoprotein endpoint rather than a lesion-specific one, and this graph already holds material on the other side of that question: it carries a PRC2 arm, an H3K27me3 entry, and analogy-ezh2-driven-mhc-silencing-in-fus-erg-and-cbfa2t3-glis2, in which two other fusions converge on EZH2-driven silencing - the opposite direction. Two convergences with opposite signs is a more interesting corpus-level fact than either one, and it says the mark is a dial fusions turn rather than a direction they all push.

For the index fusion specifically, the relevant existing claim is the polycomb phenocopy. This graph records that menin inhibition induces PRC1.1-dependent H2AK119ub deposition silencing a subset of MLL-fusion targets, and that H3K27me3 has been measured under the genetic arm of that work. So in KMT2A-rearranged disease the drug ADDS Polycomb marks to fusion targets, while both MBTD1 fusions REMOVE one. If the MBTD1 convergence holds, the comparison to make is not 'do these fusions resemble KMT2A fusions' but 'is Polycomb the axis that all of them move, in whichever direction their lesion happens to point'. That is a question about the whole corpus and it is not currently asked anywhere in it.

What would make this hypothesis uninteresting. If arm 1 returns H3K27me3 loss across all low-grade ESS regardless of fusion, and arm 2's base-rate control shows no excess depletion at fusion sites, then both arms are properties of their tissue and their comparator respectively, and the convergence is an artefact of looking at two H3K36me3-adjacent systems. That is the likeliest outcome and the entry is filed at speculative for it.

Facts and figures
statementThe chromatin endpoint both MBTD1 fusions produce is H3K27me3-poor chromatin, reached by opposite mechanisms: in ZMYND11::MBTD1 by building an H3K36me3-plus-acetylation state that excludes PRC2, and in MBTD1::EZHIP by delivering an H3K27M-mimicking peptide that poisons EZH2 catalysis. On this reading the shared endpoint is the selected phenotype and MBTD1's presence in both fusions is a coincidence of breakpoint availability rather than a mechanistic thread.
origindata-pattern
origin noteTwo published observations in two literatures that do not cite each other, noticed while writing the MBTD1 polarity comparison. Neither paper mentions the other's fusion.
prior probability0.15
prior noteLow, and the reason is a base rate rather than doubt about either measurement. H3K36me3 and H3K27me3 are mutually antagonistic genome-wide in every cell type, so the leukaemia arm's observation may be a restatement of where the fusion binds. This project has written four proportions whose null was 50% by construction and read their agreement as a pattern; this is the same shape and the prior is set accordingly.
the base rate problem stated plainlyH3K36me3 and H3K27me3 are mutually antagonistic throughout the genome, in every cell type, with no fusion required. So 'the fusion's H3K36me3-overlapping peaks are H3K27me3-devoid' may be a description of where the fusion binds rather than of anything the fusion does. The comparator that carries information is H3K36me3-matched, expression-matched, non-fusion-bound regions - not zero, and not the genome average. Every version of the leukaemia arm that omits that comparator is a proportion without a null.
why the sarcoma arm is the informative oneIts comparator is a genotype, not a chromatin state: fusion-positive against fusion-negative low-grade ESS from the same archive. That comparison cannot be confounded by the H3K36me3/H3K27me3 antagonism, the assay is validated and routine, and no measurement of any kind currently exists.
absence classmeasured-not-reported
absence class noteThe load-bearing absence is H3K27me3 in MBTD1::EZHIP tumour tissue, and it is measured-not-reported: H3K27me3 immunohistochemistry is a routine diagnostic stain on formalin-fixed material in current use for PFA ependymoma and validated in 2026 across 102 CNS germ cell tumour cases, and the fusion-positive tumours were formalin-fixed because a FISH assay was built for that material, so archival blocks plausibly exist and the stain may well have been done as part of a workup and never written down. THE LEUKAEMIA ARM IS A DIFFERENT CLASS AND IS SPLIT: comparator (i), H3K36me3-matched unbound regions in the deposited data, is measured-not-analysed and is pure compute; comparators (ii) and (iii), before fusion expression and after fusion withdrawal, are never-measured and need new material. NO GEO ACCESSION IS NAMED anywhere in this entry because none was looked up in this beat, and naming one from memory is the failure this field exists to prevent.
predictions
[
  "H3K27me3 is depleted in MBTD1::EZHIP-positive endometrial stromal sarcoma tissue relative to fusion-negative low-grade ESS, in the EZHIP pattern of broad-domain loss with CpG-island retention.",
  "Fusion-negative low-grade ESS - particularly JAZF1-SUZ12, which fuses a NuA4 component to a PRC2 CORE subunit - RETAINS H3K27me3. If the whole family is H3K27me3-low, the mark is a property of the tumour type and arm 1 is uninformative.",
  "In the deposited ZMYND11::MBTD1 occupancy data, H3K27me3 depletion at fusion peaks EXCEEDS depletion at H3K36me3-matched regions not bound by the fusion. If it does not, the leukaemia arm is a null.",
  "H3K27me3 INCREASES at former fusion sites when the fusion is withdrawn or when BET inhibition collapses the acetylated state, which is the only measurement that could show the exclusion is active rather than incidental.",
  "The convergence does NOT extend to H2AK119ub. The two mechanisms proposed act on PRC2 catalysis and on PRC2 access, neither on PRC1, so a PRC1 mark should behave differently - and if it tracks H3K27me3 exactly, the shared endpoint is a general repressive-chromatin phenomenon and not a PRC2 story."
]
experiments
[
  "exp-h3k27me3-and-ezhip-ihc-in-archival-mbtd1-ezhip-sarcoma"
]
analyses
[]
falsifiers
[
  "SCOREABLE FALSIFIER 1, the base-rate control and the one most likely to fire. In the deposited ZMYND11::MBTD1 occupancy data, compare H3K27me3 signal at fusion peaks against H3K36me3-matched, expression-matched regions NOT bound by the fusion, in the same murine cells. If the depletion at fusion peaks is less than 1.5-fold greater than at matched unbound regions, the leukaemia arm is REFUTED as a fusion-specific claim and must be reported as a restatement of the fusion's binding preference. NOTE: no GEO accession is named here because none was looked up in this beat.",
  "SCOREABLE FALSIFIER 2. H3K27me3 immunohistochemistry on archival MBTD1::EZHIP-positive endometrial stromal sarcoma, scored by the convention used for PFA ependymoma, against at least five fusion-negative low-grade ESS controls from the same archive. If fusion-positive tumours retain H3K27me3 at or above the control level, the sarcoma arm is REFUTED. If the CONTROLS are also H3K27me3-low, the result is UNINFORMATIVE rather than supportive and must be reported as such - the tumour type, not the lesion, would be the explanation.",
  "SCOREABLE FALSIFIER 3, which kills the framing rather than the observation. If the MBTD1 half of MBTD1::EZHIP contributes nothing - if a cage mutant, or EZHIP expressed from the MBTD1 promoter without MBTD1, produces the same H3K27me3 effect - then the sarcoma arm is an EZHIP overexpression event and the 'two fusion mechanisms converging' framing is REFUTED even though both observations stand.",
  "SCOREABLE FALSIFIER 4, on the corpus side and requiring no wet work. This graph holds analogy-ezh2-driven-mhc-silencing-in-fus-erg-and-cbfa2t3-glis2, in which two fusions converge on EZH2-driven SILENCING - the opposite direction - and records that menin inhibition ADDS PRC1 marks to KMT2A-fusion targets. If a survey of this corpus's fusion entries finds H3K27me3 moving in both directions about equally often, then 'fusions deplete H3K27me3' is not a pattern and this hypothesis's general form is REFUTED at corpus level.",
  "SCOREABLE FALSIFIER 5. If H2AK119ub tracks H3K27me3 exactly at ZMYND11::MBTD1 sites, the endpoint is general repressive-chromatin exclusion rather than a PRC2-specific phenomenon, and the mechanistic contrast between 'exclusion' and 'catalytic antagonism' that this hypothesis rests on loses its content."
]
How this could bite us back
  • If the leukaemia arm's base-rate control fires, this entry becomes a worked example of the error it was written to avoid, and that is the likeliest single outcome. The honest reading is that it is filed at 0.15 for that reason and should be reported as a null rather than quietly narrowed if the control fails.
  • A hypothesis that both fusions deplete H3K27me3 invites the thought that H3K27me3 restoration is therapeutic, and there is no agent. EZH2 activators do not exist, the H3K27M glioma field has pursued this far harder than a three-case sarcoma or a twelve-case leukaemia will, and nothing approved has come of it.
  • In the KMT2A-rearranged disease this project actually studies, the Polycomb direction is the opposite one: menin inhibition ADDS PRC1 marks to fusion targets, and PRC1.1 subunit loss is an established resistance route. So if anything this hypothesis argues that Polycomb is a dial fusions turn in whichever direction their lesion points, which makes it a poor source of therapeutic direction for any of them.
  • The sarcoma arm's cheapest test may not be able to read its own prediction. The expected signature is loss of broad H3K27me3 domains with retention at CpG islands, and a nuclear immunohistochemical stain integrates over the whole genome. A partial stain would be uninterpretable, and the experiment entry has to say what counts as a positive before it is run.
Still open
  1. Do fusion-negative low-grade endometrial stromal sarcomas retain H3K27me3? This is the control that decides whether the whole sarcoma arm is interpretable, it needs no fusion-positive case at all, and it can be run on any existing low-grade ESS series. It should be done FIRST, before hunting for the three fusion-positive blocks.
  2. What is the H3K27me3 status of JAZF1-SUZ12 tumours specifically? That fusion puts a PRC2 CORE subunit at the C-terminus where MBTD1::EZHIP puts a PRC2 antagonist. If both tumour types are H3K27me3-low, the mark is a property of low-grade ESS; if they differ, the family is a natural experiment in PRC2 gain versus PRC2 loss within one histology - which would be the most informative thing in this entry and nobody has looked.
  3. Does H3K27me3 return at ZMYND11::MBTD1 sites under BET inhibition? I-BET151 collapses the acetylated state and releases the differentiation block. If the mark comes back, PRC2 exclusion is active and acetylation-dependent; if it does not, the H3K27me3 absence is a property of the underlying chromatin and the fusion is incidental to it. The reagent, the model and the phenotype all already exist in the founding paper.
  4. Does this corpus's H3K27me3 evidence point in one direction across fusion families or in both? This graph already holds a case of two fusions converging on EZH2-driven SILENCING, and records menin inhibition ADDING PRC1 marks to fusion targets. A corpus survey is compute, needs no new material, and would say whether 'fusions deplete H3K27me3' is a pattern or a selection of two examples - which is falsifier 4 and the cheapest thing in this entry.
  5. Is there any H3K36me3 measurement in MBTD1::EZHIP sarcoma? If MBTD1's cage or its NuA4 membership puts the sarcoma chimera at transcribed chromatin too, then both fusions would occupy H3K36me3-marked regions and the convergence would have a shared address as well as a shared endpoint - which is a much stronger claim, and entirely unexamined.
Every source behind this entry (6)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    ZMYND11-MBTD1 binds 16,114 sites concentrated at promoter-proximal regions and gene bodies overlapping H3K36me3 and devoid of H3K27me3, recruits the NuA4/Tip60 acetyltransferase complex and establishes super-enhancers at proto-oncogenes, sustaining an active chromatin state enriched in histone acetylation and devoid of repressive histone marks.

    PMID 33594072DOI
    What this evidence cannot show

    Murine primary HSPCs with a retrovirally overexpressed, tagged human cDNA; the occupancy is on a tagged construct, not an endogenous allele. THE H3K27me3 STATEMENT IS AN OVERLAP OBSERVATION, NOT A CAUSAL ONE: fusion peaks lack the mark, and no measurement exists of H3K27me3 before fusion expression, after fusion withdrawal, or at H3K36me3-matched unbound regions. 'Refractory to PRC2' is the authors' proposal in discussion. Note also that ZMYND11[tiab] AND H3K27me3[tiab] returns 0 records, because the abstract never names the mark - rule 80 in its clearest form, and the reason this observation is easy to miss. Full text read via PMC7886901.

    Evidence that argues against this (1)
    • H3K36me3 and H3K27me3 are mutually antagonistic across the genome in every cell type, independently of any fusion, so 'H3K36me3-overlapping peaks are H3K27me3-devoid' may be a restatement of where the fusion binds and carry no information about what the fusion does. Without an H3K36me3-matched unbound comparator the observation has no null, which is the failure mode this project has documented four times in its own DepMap proportions. Devoucoux M et al. 2022
  2. E2strong 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, with the inhibitory function mapping to a short conserved region containing a methionine whose single substitution abolishes the effect and whose flanking sequence resembles the H3K27M oncohistone; the fusion protein contains the majority of MBTD1, lacking only its final 39 residues, joined to the C-terminal half of CATACOMB.

    PMID 31281901DOI
    What this evidence cannot show

    Full text read this session via PMC6609211. ECTOPIC HA-FLAG EXPRESSION IN HEK293T AND HCT-116; the authors' 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 a redistribution, and E4 shows the expected effect IS a redistribution. Genuinely new to this graph per check-citation-reuse.

    Evidence that argues against this (1)
    • Full-length CATACOMB produces the same H3K27me2/3 decrease as the fusion in this paper, so nothing establishes that the MBTD1 half contributes anything. If the fusion is functionally an EZHIP overexpression event driven off the MBTD1 promoter, then the 'convergence' in this hypothesis is between a fusion mechanism in leukaemia and a promoter-swap overexpression in sarcoma, which are not comparable objects and the hypothesis's framing collapses. Hubner JM et al. 2019
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    EZHIP/CXorf67 inhibits PRC2 through a small highly conserved C-terminal peptide that mimics K27M-mutated histones and binds the SET domain of EZH2, and the low-grade endometrial stromal sarcoma fusion family - JAZF1-SUZ12, JAZF1-PHF1, EPC1-PHF1, MEAF6-PHF1 and MBTD1-CXorf67 - consists of an N-terminal NuA4 acetyltransferase-complex component fused to a C-terminal Polycomb-group subunit, with JAZF1 newly identified as a NuA4 subunit and MBTD1 described as a NuA4 subunit gene.

    PMID 30923826DOI
    What this evidence cannot show

    THE CITATION SUPPORTS ONLY THE FIRST CLAUSE. The EZHIP/EZH2 SET-domain mechanism is this paper's, read at abstract level; the ESS fusion-family architecture and the JAZF1 and MBTD1 NuA4 assignments are PMID 31281901's, from its full text, and are recorded in this entry's E2 and in fusion-mbtd1-ezhip E2. The two are combined in one claim here for readability and that is a defect in the claim's construction, flagged rather than hidden: a reader checking this citation against the second clause will not find it. The ependymoma system has no relation to sarcoma or leukaemia.

    Evidence that argues against this (1)
    • The family rule is drawn from four or five fusions in one rare tumour type, and one of its members was assigned to NuA4 by the same paper that stated the rule - so rule and instance were established together. A 2026 review reaches the opposite subunit assignment for MBTD1, calling it PRC1-associated, which is wrong against three primary papers but shows the category membership this hypothesis leans on is not universally carried. Yu X et al. 2026
  4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    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, and cited for the SPATIAL SIGNATURE, which is what makes the proposed sarcoma experiment scoreable: the prediction is not 'H3K27me3 absent' but 'broad domains lost, CpG islands retained'. An immunohistochemical stain cannot resolve that distinction either, so a positive stain supports the hypothesis and a partial one is uninterpretable without ChIP. No fusion and no sarcoma appears in this paper. Genuinely new to this graph.

    Evidence that argues against this (1)
    • If the expected signature is a redistribution rather than a loss, then the only chromatin evidence that exists for the MBTD1 fusion - a bulk Western blot - is the wrong instrument, and so is the immunohistochemical experiment this hypothesis proposes as its cheapest test. A nuclear stain integrates over the whole genome and would read a redistribution as a partial loss or as nothing, depending on the proportion of the genome in broad domains in that cell type. Piunti A et al. 2019
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

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

    PMID 42453183DOI
    What this evidence cannot show

    Abstract-level. Cited ONLY to establish that the stains are routine, scoreable and validated on archival material, which is what makes arm 1 an afternoon rather than a project. The disease is a CNS germ cell tumour, unrelated to endometrial stromal sarcoma. NO ANTIBODY CLONE IS RECORDED because none was retrievable at abstract level, and under this project's rule 74 the clone is the identifier - so this supports 'the assay exists and performs' and not 'this reagent will work here'. Denominators differ between markers (102, 82, 97) because not every case was stained for every marker. Genuinely new to this graph.

    Evidence that argues against this (1)
    • Sensitivity measured where the lesion is EZHIP OVEREXPRESSION does not transfer to a fusion retaining only EZHIP's C-terminal half: an N-terminal-epitope antibody would miss the fusion protein entirely while performing perfectly everywhere it was validated. That makes the EZHIP arm of the proposed stain unreliable without epitope information, and leaves H3K27me3 as the only readable arm - which is also the arm E4's counter-evidence says a stain may not be able to read. Piunti A et al. 2019
  6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    Bounded literature absence, 2026-09-23: MBTD1[tiab] AND H3K27me3[tiab] returns exactly 1 record, PMID 42238311, which is a 2026 review and not a measurement; ZMYND11[tiab] AND H3K27me3[tiab] returns 0, despite the founding leukaemia paper reporting 16,114 fusion peaks devoid of the mark; and EZHIP[tiab] AND H3K27me3[tiab] returns 33, so the mark's literature is attached to the free protein and not to either fusion.

    What this evidence cannot show

    One [tiab] tag per term, never the parenthesised group form; every querytranslation read back; pubmed_safe.py raises rather than returning a zero on an empty translation. THE ZMYND11 ZERO IS THE POINT AND IS SELF-DOCUMENTING: the founding paper's own H3K27me3 result is invisible to an abstract query because its abstract does not name the mark, which is rule 80 demonstrated inside the item that cites it. The name list searched was H3K27me3 only - not 'H3K27 trimethylation', 'trimethylated H3K27', 'H3 lysine 27' or 'Polycomb', so the counts are a floor. Titles and abstracts only; supporting figures not searched.

    Evidence that argues against this (1)
    • These counts bound published titles and abstracts and say nothing about whether the measurement exists. For the sarcoma arm specifically, H3K27me3 immunohistochemistry is so routine in soft-tissue and CNS pathology that a fusion-positive case could easily have been stained as part of a diagnostic workup and reported in a table nobody indexed - which would make this measured-not-reported, as the absence class says, rather than absent. The right next step is not a better query but a request to the reporting institution. Dewaele B et al. 2014
Connected to