Oncofusion

start / graph / ana-q37-a-damage-state-resection-occupancy-dataset-does-exist-and-it-cannot-reach-the-kmt2a-locus

Analysisstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q37-archive-sweep

A genome-wide DAMAGE-STATE occupancy dataset for CtIP and MRE11 does exist and is public - GSE218590, 58 samples, +/- 4-OHT and a light-activated Cas9 time course - and it cannot answer the KMT2A question, because there is not one AsiSI site in the 90,377 bp of the gene

also called GSE218590 is the damage-state resection dataset · no AsiSI site in KMT2A · why the DIvA system cannot answer the KMT2A question

Last cycle this graph said nobody has ever measured repair proteins on damaged DNA anywhere, for any of these proteins. That is wrong - a 58-sample public dataset does exactly that - but the breaks in it are made at specific sequences, and none of those sequences occurs anywhere inside the KMT2A gene, so it still cannot answer our question.

The same thing, for a scientist

**`absence-no-damage-state-occupancy-measurement-for-any-resection-factor-at-the-kmt2a-locus` lists, in its own `what_was_NOT_tested`, that *"no damage-state dataset of any kind was searched for, in any archive, for any of the seven proteins"*. This entry runs that search, and the dataset exists.** **GSE218590** (Deshpande et al., *Nat Commun* 2023, PMID 37717054) is 58 human samples that map exactly the reaction this graph's resection cascade describes [E1]: **CtIP (RBBP8) ChIP-seq in U2OS with and without 4-OHT** to activate AsiSI-ER, and **MRE11 ChIP-seq in HEK293 at 0, 15 and 60 minutes after a light-activated Cas9 break**, plus DNA-PK-bound MRN cleavage products. Eight CtIP SRX and six MRE11 SRX appear in ChIP-Atlas under `RBBP8` and the retired symbol `MRE11A`. **So the blanket claim is refuted.** Two resection factors have public, genome-wide, damage-conditional occupancy data in human cells. The per-protein table in that entry - which covers BRCA1, CBX3, CBX1, KDM5A, BARD1, EXO1 and DNA2 - survives, because CtIP and MRE11 are not among its seven. **The entry's own id says "for any resection factor", and that framing does not survive.** **And then the dataset fails the question anyway, for a reason worth recording in full** [E2]. Both break systems are site-directed. The AsiSI recognition sequence is `GCGATCGC`. Fetched live from Ensembl GRCh38 on 2026-09-22 over **chr11:118,336,456-118,626,832**, a 290,377 bp window centred on *KMT2A*: ``` AsiSI sites in the whole 290 kb window 1 (chr11:118,611,175) AsiSI sites inside the KMT2A gene body 0 (118,436,456-118,526,832) distance from the gene 3' end to that site 84,343 bp control chr11:117,000,000-117,290,376 0 control chr11:119,000,000-119,290,376 1 ``` **There is no AsiSI site in *KMT2A*.** AsiSI resection tracks are measured in hundreds of base pairs to a few kilobases from the cut; 84 kb is not reachable. The light-activated-Cas9 arm is worse for this purpose, because it is one guide at one designed locus. **Therefore: my own peak-level negative at the KMT2A intervals in these tracks is not a negative result** [E3]. The 180-track sweep read all eight CtIP tracks (four above the 200-peak dynamic-range gate) and all fourteen MRE11 tracks (thirteen above the gate) at the seven *KMT2A* intervals and found zero peaks. **That number must not be quoted as evidence that resection does not occur at *KMT2A* in the damage state.** No break was induced there. It is an absence of testing, and this graph's own standing rule says an absence of testing is not a negative result. **What this changes in practice.** The damage-state question at *KMT2A* remains `never-measured` - but for a sharper reason than "nobody has done a damage-state experiment". The experiment class exists, is mature, is cheap, and is **site-directed**, so closing this absence does not need a new assay: it needs the existing assay pointed at this locus, by a Cas9 guide cutting inside the breakpoint cluster region. That is a materially smaller ask than the entry currently states.

Facts and figures
scriptanalyses/q37-archive-sweep-at-the-kmt2a-locus/run.py for the peak reads; the AsiSI scan was an Ensembl GRCh38 sequence fetch and a literal string search, recorded inline
statusrun
run date2026-09-22
run byQ37-archive-sweep
absence classnever-measured
absence class noteThe damage-state occupancy of any resection factor AT THE KMT2A LOCUS remains never-measured. What changes is the reason and the cost: the assay exists and is site-directed, so the missing ingredient is a guide, not a method.
what this would cost to closea Cas9 guide cutting inside chr11:118,482,422-118,489,887 in a haematopoietic cell, with the GSE218590 readouts (CtIP ChIP, MRE11 ChIP, DNA-PK-bound fragment sequencing) at 0, 15 and 60 min. The protocol, the antibodies and the analysis are published and the light-activated-Cas9 timing is already worked out; only the guide and the cell type are new.
relationship to existing entriesCloses item 4 of `absence-no-damage-state-occupancy-measurement-for-any-resection-factor-at-the-kmt2a-locus`.data.what_was_NOT_tested. It refutes that entry's title and id insofar as they generalise to 'any resection factor', and it leaves the seven-protein per-protein table standing. It also supplies the missing experimental precedent for FALSIFIER 3 of `hyp-the-menin-arm-and-the-ctcf-arm-license-consecutive-steps-of-one-resection-reaction`.
the dataset
{
  "accession": "GSE218590",
  "pmid": "37717054",
  "n_samples": 58,
  "organism": "Homo sapiens",
  "gdstype": "Genome binding/occupancy profiling by high throughput sequencing",
  "arms_identified_from_sample_titles": {
    "CtIP U2OS AsiSI": [
      "GSM7498202 CtIP GC no 4-OHT rep1 (SRX20732396)",
      "GSM7498203 rep2 (SRX20732397)",
      "GSM7498205 CtIP GC plus 4-OHT rep1 (SRX20732399)",
      "GSM7498206 rep2 (SRX20732400)",
      "GSM7498208 CtIP PC no 4-OHT rep1 (SRX20732402)",
      "GSM7498209 rep2 (SRX20732403)",
      "GSM7498211 CtIP PC plus 4-OHT rep1 (SRX20732405)",
      "GSM7498212 rep2 (SRX20732406)"
    ],
    "MRE11 HEK293 light-activated Cas9": [
      "GSM6753029/30 no light +NU-7441 rep1/2 (SRX18353285/6)",
      "GSM6753031/32 15 min post-light rep1/2 (SRX18353287/8)",
      "GSM6753033/34 60 min post-light rep1/2 (SRX18353289/90)"
    ]
  },
  "chipatlas_antigen_strings": {
    "CtIP": "RBBP8",
    "MRE11": "MRE11A - the RETIRED HGNC symbol; an exact-match sweep on MRE11 returns zero"
  }
}
the asisi scan
{
  "recognition_site": "GCGATCGC (AsiSI / SgrDI)",
  "source": "Ensembl REST /sequence/region/human, coord_system_version=GRCh38, fetched 2026-09-22",
  "window_chr11_118336456_118626832_bp": 290377,
  "asisi_sites_in_window": 1,
  "asisi_site_coordinate": 118611175,
  "asisi_sites_inside_kmt2a_gene_body_118436456_118526832": 0,
  "distance_from_gene_3prime_end_to_nearest_site_bp": 84343,
  "control_windows_same_width": {
    "chr11:117,000,000-117,290,376": 0,
    "chr11:119,000,000-119,290,376": 1
  },
  "conclusion": "the AsiSI arm induces no break anywhere in or near KMT2A, so occupancy readouts from it at this locus measure the undamaged state of a cell in which breaks exist elsewhere"
}
my own negative is uninformative and that is the point
{
  "what_was_measured": "8 RBBP8 and 14 MRE11A ChIP-Atlas hg38 peak sets read at the seven KMT2A intervals; zero peaks in the breakpoint region in every interpretable track",
  "why_it_must_not_be_quoted_as_a_negative": "no double-strand break was induced at this locus in either arm. This is the standing rule 'an absence of testing is not a negative result' applied to my own measurement, in the same entry that reports it.",
  "dynamic_range_gate": "4 of 8 RBBP8 tracks (61-688 peaks) and 1 of 14 MRE11A tracks fall below the pre-registered 200-peak gate and are recorded as uninterpretable rather than negative"
}
How this could bite us back
  • The finding cuts against my own headline in this beat. I found a damage-state dataset, read it at the locus, and had a clean negative in hand before checking whether a break could have been induced there - the same order of operations that produces every false negative in this project's registry.
  • If a future beat points a Cas9 break at the KMT2A breakpoint cluster to run this assay, it is deliberately creating the lesion that causes infant leukaemia in a human cell. That is routine in a cell line and is not routine in a primary CD34+ progenitor destined for any translational use, and the biosafety framing belongs in the protocol rather than in a footnote.
Still open
  1. Which AsiSI site in the DIvA system is the nearest cut to any recurrent leukaemia translocation breakpoint? The whole DIvA/AsiSI resection literature is a fixed panel of about 80 cut sites, and nobody appears to have asked whether any of them sits in a clinically recurrent breakpoint cluster region - a one-afternoon coordinate join with a published breakpoint table.
  2. Does the light-activated Cas9 protocol of GSE218590 work in a CD34+ progenitor or a B-lineage precursor line? The timing and the antibodies transfer; the delivery does not obviously transfer, and that is the practical obstacle to pointing this assay at KMT2A.
  3. Would DNA-PK-bound fragment sequencing at a Cas9 break inside the KMT2A breakpoint cluster distinguish the menin-arm and CTCF-arm predictions? The hypothesis it would test is already written and its falsifier already names the interval; the assay that could score it was not known to this graph until now.
Every source behind this entry (3)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    A public, genome-wide, damage-conditional occupancy dataset for two resection factors exists: GSE218590 holds 58 human samples including CtIP ChIP-seq in U2OS with and without 4-OHT activation of AsiSI-ER, and MRE11 ChIP-seq in HEK293 at 0, 15 and 60 minutes after a light-activated Cas9 break, alongside sequencing of DNA-PK-bound MRN cleavage products.

    Genome-wide analysis of DNA-PK-bound MRN cleavage products supports a sequential model of DSB repair pathway choice. · Deshpande RA, Marin-Gonzalez A, Barnes HK, et al. · Nature Communications 2023
    PMID 37717054DOI
    What this evidence cannot show

    U2OS is an osteosarcoma line and HEK293 an embryonic kidney line; neither is haematopoietic and neither is KMT2A-rearranged. The abstract was read in full from efetch XML; the figures were not opened, and the sample arms above were reconstructed from GEO sample titles rather than from the methods section. The DNA-PK-bound-fragment assay is a cleavage-product readout rather than an occupancy readout and is not the same measurement as the two ChIPs.

    Evidence that argues against this (1)
    • The seven-protein table in the entry this refutes does not include CtIP or MRE11, so its per-protein claims are untouched. Someone quoting this entry as 'the damage-state absence is closed' would be wrong: it is closed for two proteins that were never in the list, at loci that are not KMT2A. Q35-encode-reanalysis 2026
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    The dataset cannot reach the KMT2A locus, because there is no AsiSI site in the gene. A live Ensembl GRCh38 fetch of chr11:118,336,456-118,626,832 (290,377 bp) on 2026-09-22 and a literal search for the AsiSI recognition sequence GCGATCGC returns exactly one site in the window, at chr11:118,611,175, which is 84,343 bp beyond the 3' end of KMT2A, and ZERO sites inside the 90,377 bp gene body chr11:118,436,456-118,526,832. Two same-width control windows on chr11 return 0 and 1 sites, so the density is as expected and the gene is not unusual - it simply has none.

    What this evidence cannot show

    The search is for the exact 8-mer on the forward strand; AsiSI is palindromic so the reverse strand adds nothing, but methylation-sensitive cutting means only a subset of AsiSI sites are cut in any given cell line, which makes this an upper bound on cuttable sites rather than a count of them. The reference assembly is not the U2OS genome; a line-specific variant could create or destroy a site. Resection track lengths from AsiSI breaks were taken as 'hundreds of bp to a few kb' from the paper's own statement that products fall within 200 bp of the break under DNA-PK inhibition, not from an independent measurement.

    Evidence that argues against this (1)
    • If long-range resection in this system extended tens of kilobases, an 84 kb gap would still be out of reach but the argument would be weaker. The paper reports products within 200 bp of the break when DNA-PK is inhibited and 'greater dispersal' without inhibition, and does not quantify the upper tail, so the 84 kb figure is comfortable but the dispersal bound is taken from a qualitative phrase. Deshpande RA et al. 2023
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

    The peak-level reads this beat ran on those tracks are therefore uninterpretable as a negative. All 8 CtIP and all 14 MRE11 ChIP-Atlas hg38 peak sets were read at the seven KMT2A intervals and returned zero peaks at the breakpoint region, with the CTCF positive control passing and the chr2 gene-desert negative control at zero in 180 of 180 tracks - but no break was induced at KMT2A in either arm, so the zero records that nothing was tested rather than that nothing was there.

    What this evidence cannot show

    Four of the eight CtIP tracks (61 to 688 peaks) fall below the pre-registered 200-peak dynamic-range gate, so even the uninformative negative rests on four tracks and not eight. The MRE11 HeLa and LNCaP tracks in the same census are from unrelated studies with no induced break at all.

    Evidence that argues against this (1)
    • An argument that a negative is uninformative can be used to explain away any negative, which is the failure mode Q35 flagged against itself when it argued that a resting-state ENCODE negative would be uninformative. The difference here is checkable and quantitative rather than interpretive: the break-inducing enzyme has a defined recognition sequence and that sequence does not occur in the gene. Q32-brca1-hr-licensing 2026
Connected to