Oncofusion

start / graph / absence-h3k9me3-at-the-kmt2a-locus-was-assayed-and-never-reported

Analysisstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q102-h3k9-rearrangement-brake

The one experiment that would separate the H3K9me1/2 rearrangement brake from H3K9me3 was run, deposited as eight public samples, and reported nowhere in the paper

also called the missing H3K9me3 arm of GSE210480 · H3K9me3 at the KMT2A BCR: measured, not reported

The paper that found a silencing tag controls whether the KMT2A gene breaks also measured a closely related tag that should have behaved differently, and never showed the result. The data are sitting in a public archive, eight samples, and nobody has looked.

The same thing, for a scientist

PMID 37788669's own text states that it performed *'chromatin immunoprecipitation (ChIP) sequencing for H3K9me1/2/3 methylation marks in control and KDM3B siRNA transfected RPE cells'*, and its STAR Methods lists an H3K9me3 antibody (Abcam ab8898) among the four used [E1]. **No H3K9me3 result appears anywhere in the paper** - not in a main figure, not in a supplemental figure legend, not in a result sentence, not in the discussion. The only mention of me3 outside the antibody list is an in vitro enzyme blot in Supplemental Figure 2A [E1]. **The data are public:** GSE210480 contains **eight H3K9me3 ChIP-seq samples** - control, siKDM3B, siG9A and siKDM3B+siG9A, two replicates each - alongside matched H3K9me1, H3K9me2, CTCF and Input, 60 samples total [E2]. This is a `measured-not-analysed` absence in the strongest sense the vocabulary allows: the assay was run, the files are deposited, the analysis costs compute and no wet work. **And it is load-bearing for two entries.** `mark-h3k9me3` argues H3K9 methylation restrains the KMT2A-fusion programme and warns that lowering it may worsen the disease; this beat's `mechanism-h3k9me1-2-ctcf-balance-gates-kmt2a-rearrangement` argues raising H3K9me1/2 promotes rearrangement and that lowering it is protective. **Mark identity is the proposed resolution, and the missing eight samples are the measurement that tests it.**

Go deeper
What is present, what is absent, and how the absence was established

Present in the paper.

  • H3K9me1 ChIP-seq: genome-wide MA-style plots, the KMT2A BCR profile, the CTCF-site quantification, the siG9a rescue, the double-knockdown reset, the 88/109 Mb genome-wide rescue figure.
  • H3K9me2 ChIP-seq: the genome-wide increase, the BCR-flank profile, the 219/292 Mb rescue figure.
  • CTCF ChIP-seq and ChIP-qPCR.
  • One H3K9me3 western blot of an in vitro demethylation reaction with recombinant KDM3B (Supplemental Fig. 2A) - an enzyme-activity control, not a chromatin measurement.

Absent from the paper. Any H3K9me3 ChIP result, in any arm, at any locus.

How this was established. The full text and all supplemental figure legends were retrieved and the stripped body searched for every occurrence of me3 and K9me3 1. Two hits: the STAR Methods antibody list, and the Supplemental Fig. 2A legend. No third. The methods sentence saying me1/me2/me3 ChIP-seq was performed is in the Results, three sentences before the first me1 result.

Why this is not an accusation. Depositing an assayed arm and not reporting it is ordinary practice when the result is uninformative - and 'H3K9me3 did not change' would be exactly the uninformative result that a reader can reasonably guess at. But a guess is not a measurement, and this project's rule is that the reader must not write what they think the paper probably says. What is on record is that the assay exists, the files exist, and the answer is unstated.

Why it decides something, in both directions
if H3K9me3 does NOT change at the KMT2A BCRif H3K9me3 DOES change with me1/me2
the brake is genuinely me1/me2-specificthe brake is a general H3K9-methylation effect
mark identity resolves the tension with mark-h3k9me3 cleanlythe tension is real and unresolved: the same mark restrains the fusion's targets and destabilises its gene
G9a's annotated me1/me2-only specificity is the mechanistic reason, and a G9a-selective inhibitor is the right toolthe GLP/EHMT1 half of UNC0642 - annotated as an me3 writer - may be doing part of the work, and the dual inhibitor is the wrong tool for the question
drug-unc0642's bite-back about mark-h3k9me3 softens, because the protective intervention is not touching the restraining mark at the fusion's targetsthe bite-back sharpens into a genuine contraindication

Either answer is worth having, which is the definition of an experiment worth running, and this one has already been run.

Facts and figures
absence classmeasured-not-reported
what is absentany reported H3K9me3 ChIP-seq result from PMID 37788669, in any knockdown arm, at any locus - and in particular at the KMT2A breakpoint cluster region and its exon-11 CTCF site
what closes itcompute only - download the eight H3K9me3 fastq/bigwig sets plus inputs, align to hg38, and quantify over the KMT2A BCR and the exon-11 CTCF site with the me1 and me2 arms as internal positive controls. No wet work, no cohort, no permissions. See ana-gse210480-h3k9me3-at-the-kmt2a-bcr.
archive check runtrue
archive queriedNCBI GEO (esearch db=gds) - the matching archive for an occupancy question, per this project's standing rule that no absence is bounded by literature search alone
why it is load bearingmark identity is the proposed resolution of a direct conflict between mark-h3k9me3 (H3K9 methylation restrains the fusion programme; lowering it is dangerous) and mechanism-h3k9me1-2-ctcf-balance-gates-kmt2a-rearrangement (H3K9me1/2 promotes rearrangement; lowering it is protective). The unreported arm tests whether the two marks really do behave differently at this locus.
detection methodfull text retrieved via efetch db=pmc (PMC10591855) after Europe PMC fullTextXML returned HTTP 500; <body> confirmed present; tags stripped before entity unescaping; regex /K9me3|me3/ over the 80,638-character body returned exactly 2 hits - the STAR Methods antibody list and a Supplemental Fig. 2A in vitro enzyme blot legend
bound on the detectionThis establishes that no H3K9me3 ChIP result is DESCRIBED in the retrievable text. A result present only as an unlabelled figure panel, or only in a supplementary data file not carried in the PMC author manuscript, would not be seen.
generalisable lessonA deposited-but-unreported assay arm is a distinct and cheap class of finding, and nothing in this project looks for it. The signature is a methods sentence or antibody list naming an assay that no result sentence mentions - detectable by diffing the antibody/assay list against the figure legends. That diff is worth running across this corpus's load-bearing papers.
absence class noteMIGRATED 2026-09-23. This field previously held the whole verdict as free text: 'measured-and-now-analysed-for-RPE; never-measured in any haematopoietic cell'. That string was out of the controlled vocabulary, so a checker could not read it - but it said more than any token can, and it has been kept here rather than deleted. Title: the experiment 'was run, deposited as eight public samples, and reported nowhere' - so the data exists and the number was never written down, which is exactly measured-not-reported.
what exists
{
  "accession": "GSE210480",
  "organism": "Homo sapiens (RPE)",
  "total_samples": 60,
  "h3k9me3_samples": 8,
  "h3k9me3_sample_titles": [
    "CTRLsiKDM3B_H3K9me3_rep1",
    "CTRLsiKDm3B_H3K9me3_rep2",
    "siG9A_69_H3K9ME3_rep1",
    "siG9A_70_H3K9ME3_rep2",
    "siKDM3B_43_H3K9me3_rep1",
    "siKDM3B_58_H3K9ME3_rep2",
    "siKDM3B_43_siG9A_69_H3K9ME3_rep1",
    "siKDM3B_58_siG9A_70_H3K9ME3_rep2"
  ],
  "matched_arms_also_deposited": "H3K9me1, H3K9me2, CTCF and Input in the same four conditions",
  "antibody": "Abcam ab8898 (rule 74)"
}
what closes it costed cycle 57
{
  "the_files_are_processed_not_raw": "Each of the eight H3K9me3 samples carries a per-sample **bigWig** supplementary file (e.g. GSM6430988_CTRLsiKDM3B_H3K9me3_rep1.bigWig). So this does NOT require alignment from FASTQ - it requires reading coverage over an interval, which is a much smaller job than 'compute only' implied.",
  "but_the_box_cannot_read_them": "pyBigWig FAILS TO BUILD on this Windows host ('Failed to build pyBigWig when getting requirements to build wheel' - it needs a C toolchain with zlib/curl headers), deeptools is absent, and the UCSC binaries bigWigSummary and bigWigAverageOverBed are not on PATH. numpy, scipy and pandas ARE available.",
  "why_a_hand_rolled_parser_was_refused": "The bigWig/BBI format needs header, chromosome B-tree and R-tree index parsing plus zlib block decompression before a single value is returned, and a coordinate-mapping error would produce a confident, plausible, WRONG signal value at the breakpoint cluster - which is the exact failure mode this project has hit repeatedly (a truncated DepMap payload reporting a full table off 2% of the data; probe counts computed with no expression floor). A load-bearing biological claim is the wrong place to debut an unvalidated binary parser.",
  "what_would_make_it_safe": "Any ONE of: (a) pyBigWig installed on a Linux host - the Explorer cluster would do it in one command; (b) the UCSC bigWigSummary binary on PATH; (c) a hand-rolled reader VALIDATED first against a positive and negative control locus - H3K9me3 should be high over a ZNF cluster or pericentromeric region and low at a housekeeping promoter - before any KMT2A-locus number is reported. Option (c) is acceptable only with the controls reported beside the result.",
  "status": "NOT a literature gap and NOT a data gap. It is a tooling gap on one machine, stated precisely so the next cycle can close it in minutes rather than rediscovering it."
}
CLOSED cycle 59
{
  "status": "CLOSED for RPE at 1 kb resolution. The arm was read.",
  "the_answer": "H3K9me3 does not move at the CTCF element where H3K9me1 moves 3.74x. All four me3 arms lie inside the control replicate spread and the me3 log response is 1.1% of me1's. So mark identity IS available as the resolution of the mark-h3k9me3 / mechanism-h3k9me1-2 conflict - at this element, in epithelium.",
  "but_not_over_the_whole_BCR": "me3 rises 1.34x over the 7,466 bp BCR core, replicate-consistently - and EQUALLY under siG9A (1.38x), returning to control in the double knockdown. That is the wrong shape for a KDM3B-versus-G9a balance. A quarter of the apparent rise is a chromosome-wide shift: chr11's median me3 log2FC is +0.114 against +0.004 for me1.",
  "what_replaces_this_absence": "a never-measured CELL-TYPE absence: no H3K9me1/2/3 measurement at this locus exists in any haematopoietic cell, let alone a KMT2A-rearranged one. RPE is retinal pigment epithelium.",
  "TWO_COSTINGS_IN_THIS_ENTRY_WERE_WRONG": {
    "the_inputs_do_not_exist": "this entry's what_closes_it says 'download the eight H3K9me3 fastq/bigwig sets PLUS INPUTS'. The Input samples in GSE210480 have NO supplementary file at all - verified by the king across all four Input accessions - and the deposited bigWigs are already deepTools ChIP-over-Input RATIO tracks. Read depth, duplicate rate and FRiP are therefore unrecoverable, so the depth check this entry asks for cannot be done at any price.",
    "and_the_resolution_was_never_checked": "every H3K9 track in the series is 1,000 bp binned. Verified independently by the king: run lengths over a 20 kb window are modal and median 1000, the 147 bp canonical exon-11 interval returns ONE distinct value, and its mean is identical to a 501 bp interval containing it. No sub-kilobase histone statement is obtainable. NEITHER of the two prior cycles that costed this dataset checked this."
  },
  "and_a_ratio_track_is_not_comparable_with_a_coverage_track": "this beat's ~4.7x element CTCF cannot be compared with the ~50x measured from Canela's raw coverage in another beat. Ratio tracks compress dynamic range. Position only."
}
How this could bite us back
  • This entry's own conclusion could soften a warning that should not be softened. If the me3 arm turns out null, the temptation will be to relax `drug-unc0642`'s and `mark-h3k9me3`'s contraindication - but a null me3 result in retinal pigment epithelium says nothing about H3K9me3 at fusion target promoters in a leukaemic blast, which is where that warning lives. One measurement, one cell type, one locus.
  • A re-analysis that finds what it wants to find is the failure mode here. The me1 and me2 arms must be reprocessed through the same pipeline as internal positive controls, and the pipeline must reproduce the published me1 BCR result before its me3 answer is believed.
Still open
  1. Does H3K9me3 change at the KMT2A breakpoint cluster region on KDM3B depletion, G9a depletion, or both? Eight public samples answer it with compute alone, and the answer decides whether this beat's mechanism is mark-specific. **COSTED, cycle 57:** the eight samples are deposited as per-sample bigWigs, so this is an interval-query job rather than an alignment job - but pyBigWig fails to build on this Windows host, deeptools is absent and the UCSC binaries are not on PATH. A hand-rolled BBI parser was deliberately refused for a load-bearing claim without control loci. Closable on a Linux host in one command; see data.what_closes_it_costed_cycle_57.
  2. Does H3K9me3 change genome-wide in the same data? The paper reports genome-wide skews for me1 and me2 with megabase rescue figures; the same analysis on the me3 arm would show whether the KDM3-G9a balance touches the trimethyl state at all, which bears on whether a G9a/GLP inhibitor is me3-sparing.
  3. How many other load-bearing papers in this corpus name an assay in their methods that no result sentence mentions? The detection is a diff between the antibody or assay list and the figure legends, and nothing in this project has ever run it.
Every source behind this entry (3)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    PMID 37788669 states in its Results that it performed ChIP sequencing for H3K9me1/2/3 methylation marks in control and KDM3B siRNA transfected RPE cells, and its STAR Methods lists H3K9me3 antibody Abcam ab8898 alongside H3K9me1 ab8896-100, H3K9me2 ab1220 and CTCF clone D31H2; yet no H3K9me3 ChIP result appears in any figure, supplemental figure legend, result sentence or discussion passage in the paper - the only other occurrence of the string is a Supplemental Fig. 2A in vitro enzyme western blot with recombinant KDM3B.

    Epigenetic balance ensures mechanistic control of MLL amplification and rearrangement. · Gray ZH; Chakraborty D; Duttweiler RR; Alekbaeva GD; Murphy SE; Chetal K; Ji F; Ferman BI; Honer MA; Wang Z; et al. · Cell 2023
    PMID 37788669DOI
    What this evidence cannot show

    **Established by retrieving the full text via efetch db=pmc on PMC10591855 (Europe PMC fullTextXML returned HTTP 500; a <body> element was confirmed) and regex-searching the tag-stripped 80,638-character body for /K9me3|me3/, returning exactly two hits.** Two limits on that method, both stated because they bound the claim. **(1) The PMC author-manuscript body may not contain every supplemental file.** Supplemental figure legends are present in the retrieved body - the search found and read them - but a supplemental data file, a supplementary table, or a figure panel referenced only inside an image would not be searchable this way. **(2) A tag-strip search cannot see a result presented only as an image.** If an H3K9me3 track appears in a figure panel with no legend mention, this method would miss it. So the honest claim is: no H3K9me3 ChIP result is DESCRIBED in the retrievable text of this paper. That is weaker than 'the paper contains no such result', and it is what is asserted.

    Evidence that argues against this (1)
    • Not reporting an assayed arm is ordinary and usually means the result was null, so the most likely content of the missing arm is 'H3K9me3 did not change' - which would support the paper's own me1/me2-specific framing. An absence of reporting is weak evidence of anything, and this entry's value is that the answer is cheaply obtainable, not that its concealment is suspicious. Gray ZH; Chakraborty D; Duttweiler RR; Alekbaeva GD; Murphy SE; et al. 2023
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset

    GEO series GSE210480 (Homo sapiens, 60 samples) contains eight H3K9me3 ChIP-seq samples - CTRLsiKDM3B_H3K9me3 rep1 and rep2, siG9A_69_H3K9ME3_rep1 and siG9A_70_H3K9ME3_rep2, siKDM3B_43_H3K9me3_rep1 and siKDM3B_58_H3K9ME3_rep2, and siKDM3B_43_siG9A_69_H3K9ME3_rep1 and siKDM3B_58_siG9A_70_H3K9ME3_rep2 - alongside matched H3K9me1, H3K9me2, CTCF and Input samples in the same four arms, and it is the only GEO series returned by the query H3K9me2 AND KMT2A.

    What this evidence cannot show

    A sample-title listing establishes that the assay was run and files were submitted; it does not establish sequencing depth, library quality, or that an H3K9me3 analysis would be informative. **Two practical hazards for whoever runs it.** (1) Sample titles are inconsistently cased and spelled within the series - H3K9ME3, H3K9me3, CTRLsiKDm3B - so a naive string-matched pipeline will silently drop arms; match case-insensitively and print what was matched. (2) This project has previously found a well-formed, correctly named, EMPTY supplementary archive, and a GEO SuperSeries that denied having raw data in three different ways, all three wrong. The presence of eight sample records is a strong prior that the analysis is cheap, not a guarantee. Also note the series title differs from the published title, so a title-matched join between paper and data fails.

    Evidence that argues against this (1)
    • A genome-wide dataset may be the wrong resolution for the locus question: the breakpoint cluster region is 8.3 kb and the paper's genome-wide analyses are binned at 10 kb, so a single bin cannot separate the BCR from its flank - which is exactly the distinction that separates the H3K9me1 result from the H3K9me2 result. The re-analysis must be done at read-level resolution over the locus, not from the binned tables. Gray ZH; Chakraborty D; Duttweiler RR; Alekbaeva GD; Murphy SE; et al. 2023
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    The question the missing arm answers is live in this graph and is currently answered in opposite directions by two entries: mark-h3k9me3 records that SETDB1-deposited promoter H3K9me3 represses Hoxa9 and Meis1 and that MLL-rearranged leukaemia stem cells carry LESS H3K9me3 than their normal counterparts, concluding that H3K9 methylation restrains the fusion programme and that a G9a or SETDB1 inhibitor is a candidate for making the leukaemia worse; and its own open questions name 'which H3K9 methylation state actually matters at fusion targets - me2 or me3' as unresolved.

    What this evidence cannot show

    This item summarises an existing entry rather than re-reading its primary sources, which is the correct provenance to state. **`mark-h3k9me3` may be entirely correct as written and this entry does not dispute it** - the two entries concern different marks, different DNA (fusion target promoters versus the KMT2A gene body) and different disease stages (established leukaemia versus a normal cell). The point is narrower: the resolution offered for the apparent conflict is mark identity, and mark identity is exactly what the unreported arm would confirm or refute.

    Evidence that argues against this (1)
    • Mark identity may not be sufficient even if H3K9me3 is unchanged in RPE, because the two entries also differ in cell type, locus and disease stage - so a null me3 result at the KMT2A locus in epithelium would remove one explanation for the conflict without establishing that the remaining ones are correct. Godfrey L; Crump NT; Thorne R; Lau IJ; Repapi E; Dimou D; et al. 2019
Connected to