Oncofusion

start / graph / exp-q116-the-h3k9-module-at-the-kmt2a-promoter-read-in-a-kmt2a-rearranged-cell-with-an-input-arm

Experimentmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q116-h3k9-axis-at-the-locus

The experiment this beat could not do: G9a, HP1a, KAP1 and H3K9me3 at the KMT2A promoter and exon-11 anchor in a KMT2A-rearranged B-lineage cell and a CD34+ progenitor, WITH an input arm, an IgG arm and a same-cell CTCF arm

also called the input track nobody deposited · the H3K9 module in the right cell type · what would make any of this transferable

Everything this project knows about these silencing proteins at the KMT2A gene comes from breast and blood-cancer cell lines that never get this leukaemia. The experiment that would make it relevant is a standard one and nobody has done it.

The same thing, for a scientist

**Why this is the experiment.** Across Q110, Q112, Q114 and this beat, the H3K9 machinery has now been read at the *KMT2A* locus in **K562, GM12878, SU-DIPG13, SU-DIPG36, RPE and MCF-7**. **Not one is *KMT2A*-rearranged and not one is a CD34+ or fetal progenitor.** Every occupancy statement the corpus holds about this axis at this locus is a statement about a cell that does not get this disease [E1]. **And this beat found two specific things that only the right cell type can resolve.** First, **G9a/EHMT2, HP1a/CBX5 and KAP1/TRIM28 all read at or above the 87th percentile of the 2,000 highest-expressed genes PROMOTERS in MCF-7, which is the null that survives the panel baseline**, and KAP1 carries a called HOMER peak there overlapping by 267 bp the ENCODE TRIM28 peak Q110 found in K562. Second, **the panel baseline control says this may be nothing**: every PROTEIN antigen in the deposit reads 95.8-99.8 at that promoter against CTCF anchors, panel median 96.87, which is what an active promoter does to any ChIP [E2]. **The design, and the three arms every existing deposit lacks.** | arm | why no existing deposit has it | |---|---| | **input / sonicated chromatin** | neither GSE139196 nor GSE225028 deposits one. Without it, a promoter percentile cannot be separated from accessibility, copy number or the hyper-ChIPable-promoter artefact - which is precisely the ambiguity this beat could not resolve | | **IgG** | GSE225028 has none and GSE139196 states no clone for any target. Every occupancy number in this corpus from either deposit carries an unverified specificity assumption | | **same-cell CTCF** | the C4 instrument control has now FAILED in seven of seven cell types, always because a fixed external 147 bp window is scored against each cell's own summit-centred windows. A CTCF arm in the same cells, re-centred on that cell's own summit, is the only way C4 becomes a passable control rather than a standing failure | **Cells.** SEM and RS4;11 (*KMT2A*::AFF1, the index fusion), THP-1 (*KMT2A*::MLLT3), REH as a *KMT2A*-germline B-lineage comparator, and CD34+ cord-blood progenitors as the normal counterpart. **Antigens.** EHMT2/G9a, CBX5/HP1a, TRIM28/KAP1, SUV39H1, SETDB1, plus H3K9me1, H3K9me2 and H3K9me3 - because **G9a's actual products, me1 and me2, are in no deposit this project has ever read at this locus**, and the whole argument turns on which methylation state is there. **Read-out geometry, fixed in advance.** The seven stored intervals; the null is that cell's own CTCF anchor set at each interval's own width; the promoter statistic is the percentile among the 2,000 highest-expressed genes' promoters in the same cells, not among CTCF anchors - the substitution this beat had to make mid-run when its panel-baseline control fired.

Go deeper
Cost, and the cheaper subset that answers most of it

Full design: 8 antigens x 5 cell types x 2 replicates plus input and IgG = ~100 CUT&RUN libraries. cost_tier: high, and it needs a CD34+ source.

The 20% that answers 80%, cost_tier: med: SEM and CD34+ only, four antigens (G9a, HP1a, KAP1, H3K9me2), two replicates, plus input, IgG and CTCF. 2 x 4 x 2 + 3 x 2 x 2 = 28 libraries. That version decides:

  1. whether the promoter module reproduces in a KMT2A-rearranged cell at all;
  2. whether it survives an input arm - the control that would have settled this beat's central ambiguity;
  3. what the me2 state of the promoter and the exon-11 anchor is, which no deposit anywhere has ever measured at this locus.

What would make it unnecessary. If somebody deposits an input track for GSE139196 or an H3K9me2 ChIP in any KMT2A-rearranged line, the priority drops sharply. Neither exists as of a GEO sweep on 2026-09-23.

Facts and figures
questionIs the G9a / HP1a / KAP1 module at the KMT2A promoter real and present in a KMT2A-rearranged cell, and what is the H3K9me1/me2 state of the promoter and the exon-11 anchor there?
designCUT&RUN or ChIP-seq, 2 biological replicates, WITH input, IgG and same-cell CTCF arms
intervalsthe seven stored in analyses/q112-signal-level-reread/PREREGISTRATION.md
nullthat cell type's own CTCF anchor set at each interval's own width, PLUS the promoters of the 2,000 highest-expressed genes in the same cells for any promoter claim
cost tierhigh for the full design; med for the 28-library SEM-plus-CD34+ subset
minimal versionSEM and CD34+ only; G9a, HP1a, KAP1 and H3K9me2; 2 replicates; plus input, IgG and CTCF = 28 libraries
readoutpercentile of same-cell-type CTCF anchors at each interval's own width, percentile among active-gene promoters for any promoter claim, and the input-normalised fold at each of the seven intervals
cell typeKMT2A-rearranged B-lineage and myeloid lines, plus normal CD34+ progenitors - the two contexts no existing deposit covers
cell stateunperturbed, plus a TOP2-poison arm if budget allows, because both the me1/me2 mechanism and the damage-recruitment hypothesis predict the state matters
epigenetic readoutoccupancy of the H3K9 writer, reader and scaffold layer plus all three H3K9 methylation states at the KMT2A locus
shown for fusionNO - this experiment exists because nothing in this axis has ever been shown in any KMT2A fusion
absence classnever-measured
absence class noteFor the KMT2A-rearranged and CD34+ contexts no instrument has been pointed at this question in any archive; it needs wet work, not re-analysis. The MCF-7 and K562 contexts were measured-not-analysed and are now closed.
panel median provenanceC5/C5b PANEL MEDIANS ARE QUOTED AT THE FINAL 28-TRACK COUNT (2026-09-23). Earlier drafts of this entry quoted them at intermediate counts (12, 18 or 25 tracks) while the run was still going. The gene-body panel median moved 61.80 -> 61.39; the PROMOTER panel median moved 82.36 -> 96.87, i.e. UP, and the direction matters: at 96.87 the CTCF-anchor percentile at the KMT2A promoter is even less informative than the first draft said, and the promoter claim rests ENTIRELY on the active-promoter null of deviation D9.
cells
[
  "SEM (KMT2A::AFF1)",
  "RS4;11 (KMT2A::AFF1)",
  "THP-1 (KMT2A::MLLT3)",
  "REH (KMT2A germline B-lineage comparator)",
  "CD34+ cord-blood progenitors"
]
antigens
[
  "EHMT2/G9a",
  "CBX5/HP1a",
  "TRIM28/KAP1",
  "SUV39H1",
  "SETDB1",
  "H3K9me1",
  "H3K9me2",
  "H3K9me3"
]
required arms no existing deposit has
[
  "input / sonicated chromatin",
  "IgG",
  "same-cell CTCF"
]
what it decides
[
  "whether the KMT2A-promoter H3K9 module reproduces outside epithelium and erythroleukaemia",
  "whether it survives an input arm, which is the control that would have resolved this beat's central ambiguity",
  "the H3K9me2 state of the exon-11 anchor, which has never been measured in any KMT2A-rearranged cell",
  "whether C4 can pass anywhere, once CTCF is measured in the same cells and re-centred on that cell's own summit"
]
C5 final 28 track values
{
  "panel_median_kmt2a_gene_percentile": 61.39,
  "panel_median_kmt2a_promoter_percentile": 96.87,
  "per_antigen_median_promoter_percentile_vs_CTCF_anchors": {
    "TRIM28": 99.8,
    "CBX5_mut4E": 99.75,
    "KDM4B": 99.37,
    "KDM4C": 98.47,
    "CBX5_mut4A": 96.76,
    "CBX5": 96.65,
    "EHMT2": 95.83,
    "H3K9me3": 29.43
  },
  "reading": "every PROTEIN antigen reads 95.8-99.8 at the KMT2A promoter against CTCF anchors and only the histone mark (29.43) does not - so that statistic cannot discriminate between proteins at all, and every promoter claim in this beat rests on the ACTIVE-PROMOTER null instead"
}
How this could bite us back
  • If this experiment finds the module present in SEM, the therapeutic reading is a G9a/GLP inhibitor or a KAP1 disruptor acting at the KMT2A promoter of a cell that already carries the fusion - which would change transcription of the wild-type allele and of the fusion allele's 5' portion at once, in opposite directions for the disease. Nothing in the design distinguishes those two, and a promoter-directed intervention at a rearranged locus is the case where allele-resolved readout stops being optional.
  • A negative result here is the more likely outcome and would be worth the cost, because it would close an axis this project has now spent four beats on. It should be commissioned as a decision experiment with the negative pre-agreed as publishable, not as a fishing expedition.
Still open
  1. Has anybody deposited an EHMT2, CBX5 or TRIM28 ChIP in a KMT2A-rearranged line? Q114 ran a complete archive census for SUV39H1 and found two deposits; no equivalent census exists for these three, and it is a one-hour job that should precede any wet work.
  2. Would an input track for GSE139196 alone answer most of this? The submitters generated one - HOMER findPeaks used input for the ChIRC-seq arm, per the deposit's own processing notes - and did not deposit it for the ChIP arm. One email might be cheaper than 28 libraries.
  3. Is the CD34+ arm feasible at CUT&RUN scale with the cell numbers a cord-blood unit yields? The antigens here are low-abundance chromatin proteins and this entry has not costed the input material.
Every source behind this entry (2)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

    Every cell type in which any H3K9 writer, reader or eraser has been read at the KMT2A locus - K562, GM12878, SU-DIPG13, SU-DIPG36, RPE and MCF-7 - is non-KMT2A-rearranged and non-progenitor.

    What this evidence cannot show

    This is a statement about what this corpus has read, not a complete archive census. Q114 ran such a census for SUV39H1 specifically and found two deposits; no equivalent census exists for EHMT2, CBX5 or TRIM28, so a KMT2A-rearranged deposit for one of those could exist unfound. That census is the cheap step that should precede commissioning anything.

    Evidence that argues against this (1)
    • GM12878 is a B-lymphoblastoid line and is the closest lineage match any archive offers for this machinery, as Q110's own entry says. An argument could be made that the B-lineage question is partly answered there - for CBX5 at least, where the answer was zero called peaks in the gene or within 50 kb. Commissioning new work should account for that partial answer rather than describing the field as empty. source
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    In MCF-7, EHMT2/G9a reads the 87.55th-97.76th, CBX5/HP1a the 96.32nd-96.97th and TRIM28/KAP1 the 99.70th-99.91st percentile of 56,171 same-cell-type CTCF summit windows at the KMT2A promoter - while the panel median across all C1-passing tracks of all antigens at that same interval is 96.87, so the promoter is ChIP-hot for the whole panel and no single antigen's number there is specific on its own.

    GSE139196 read at the KMT2A locus, analyses/q116-h3k9-axis-at-the-locus/results_scored.json, controls C5 and C5b · Q116-h3k9-axis-at-the-locus · oncofusion-knowledge-graph / NCBI GEO 2026
    What this evidence cannot show

    The C5b panel baseline is an EXTENSION of a control pre-registered on the gene-body interval only, declared as deviation D8 and labelled as an extension. The panel median moves as tracks are added, and is quoted here at the track count stated in results_scored.json. MCF-7, one cell type, no input arm - which is the whole point of this entry.

    Evidence that argues against this (1)
    • A panel median of 96.87 is itself pulled down by the six H3K9me3 tracks, which span 0.00-77.17 at the promoter. Among the PROTEIN tracks alone the promoter reads 87.6-99.9 with no exceptions, which makes the artefact reading stronger, not weaker - and means the KAP1 peak-level call and the Q110 K562 replication are the only parts of the promoter story that do not depend on a signal percentile. source
Connected to