Oncofusion

start / graph / ana-crispr-chip-fusion-identity-resolved

Analysisestablished confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.Q-partner-provenance

The CRISPR-ChIP H3K79 partition is a mouse MLL-AF9 result: resolved from GEO after the full text proved unreachable

also called which fusion did CRISPR-ChIP use · GSE192544 provenance · menin-dependent DOT1L unit provenance

The screen this project treats as its strongest independent support - the one splitting the H3K79 chromatin tag into a fusion-driven half and a normal half - was run in mouse leukaemia cells carrying MLL-AF9, not the infant fusion MLL-AF4. We could not read the paper, so we read the raw data records it deposited, and every single H3K79 experiment in them is mouse MLL-AF9.

The same thing, for a scientist

Cycle 48 left an open question: which MLL translocation did Gilan et al. use for the H3K79me2 CRISPR-ChIP screen? The published abstract does not say and the full text is unreachable (see E1 caveats). The question is answerable without the paper, because the study's sequencing is public. GEO SuperSeries **GSE192562** holds three subseries. In the CRISPR-ChIP subseries **GSE192544**, the H3K79me2 screen samples (GSM5751161 'MLLAF9 screen control'; GSM5751162 'MLLAF9 H3K79me2 CRISPR-ChIP day 9', antibody ab3594) are annotated `organism: Mus musculus`, `source: AML`, `cell type: MLL-AF9`, while the H3K4me3 and Pol II arms are human K562 (CML) [E2]. In the validation subseries **GSE192555**, **all ten** H3K79me2-bearing samples - control guide, *MLLT10* KO guides 1 and 2, *DOT1L* KO guides 1 and 2, DMSO, menin-inhibitor-treated, DOT1L-inhibitor-treated, and the two FLAG-MLLAF9 occupancy samples - are `Mus musculus`, `cell type: MLLAF9`; the thirteen human samples in that series are K562 and carry only Pol II, H3K4me3 and H3K27ac [E3]. The RNA-seq subseries **GSE192561** is 24 of 24 `Mus musculus`, source AML [E4]. **The menin-dependent oncogenic DOT1L unit, the MLLT10-directed endogenous unit, and the menin-inhibitor H3K79me2 experiment are therefore all murine MLL-AF9 (KMT2A::MLLT3) myeloid results.** No human sample contributes to any H3K79 conclusion and KMT2A::AFF1 appears nowhere in the deposited data.

Go deeper
What was actually deposited
SubseriesAssaySamples bearing on H3K79OrganismCell type
GSE192544CRISPR-ChIPGSM5751161-2Mus musculusMLL-AF9 (AML)
GSE192544CRISPR-ChIPGSM5751157-60Homo sapiensK562 (CML) - H3K4me3/Pol II only
GSE192555ChIP-seqGSM5751582-91 (10 of 10)Mus musculusMLLAF9
GSE192561RNA-seq24 of 24Mus musculusAML

The shared treatment-protocol string in GSE192544 reads: "K562 and MLLAF9 Cas9 cells (100x10^6 per cell line) were infected with the pooled lentiviral sgRNA libraries at a multiplicity of infection of 0.3 ... ChIP performed at day 6 (H3K4me3) and day 9 (H3K79me2)." That is the study's entire cell-line roster 2.

Why this matters more than an ordinary provenance note

Cycle 48 built hyp-menin-inhibition-spares-endogenous-h3k79-unit on this result and named the missing fusion as "the fact this hypothesis most needs". The answer is the unfavourable one. The partition is not a demonstration in the index fusion; it is a demonstration in the partner this project has already imported from repeatedly.

That does not make the partition wrong for KMT2A::AFF1. It makes it untested there, and it changes the entry's standing from strongest unbiased support for the menin interface to strongest unbiased support obtained in a different fusion, in a different species, in a different lineage. Three differences stack: MLL-AF9 versus MLL-AF4, mouse versus human, myeloid versus B-lymphoid.

It also converts one of that hypothesis's own falsifiers into a live experiment rather than a worry: "the partition cannot be demonstrated in a KMT2A::AFF1 model" is now the obvious next measurement (exp-crispr-chip-h3k79-partition-in-human-kmt2a-aff1).

The specific asymmetry that makes MLL-AF9 a poor proxy for this claim

MLL-AF9 carries the AF9 ANC1 homology domain (AHD), a disordered module that binds DOT1L directly 5. KMT2A::AFF1 carries no AHD: UniProt P51825 cross-references exactly three Pfam families - PF05110 (AF-4), PF18875 (AF4_int) and PF18876 (AFF4_CHD) - and not PF17793 (AHD), which both P42568/MLLT3 and Q03111/MLLT1 do carry 6.

So in MLL-AF9 the preprint's phrase "the malignant driver directs the catalytic activity of DOT1L" has a known molecular meaning. In KMT2A::AFF1 it has none yet - see mechanism-kmt2a-aff1-dot1l-recruitment-route. A menin-dependent oncogenic DOT1L unit is precisely the kind of claim whose transferability turns on the interface that is missing.

The method, because it generalises

When a paper's system is not stated in its abstract and the full text is paywalled, the sequencing deposition usually states organism and cell type per sample and is free. esearch db=gds, then the GEO accession-display text endpoint (acc.cgi?acc=<GSE>&targ=all&form=text&view=brief), gives !Sample_organism_ch1, !Sample_characteristics_ch1 = cell type: ... and the growth and treatment protocols. This beat used it after four blocked attempts at the publisher and the preprint. It is a bounded-unreadable escape hatch that this project did not previously have written down.

Facts and figures
scriptnone - read directly from the NCBI GEO accession-display text endpoint
statusrun
result summaryEvery H3K79me2 experiment in the CRISPR-ChIP study - the screen, the MLLT10 and DOT1L knockouts, and the menin-inhibitor treatment - is a murine MLL-AF9 (KMT2A::MLLT3) AML sample. 0 of 10 H3K79me2 ChIP-seq samples and 0 of 2 CRISPR-ChIP H3K79me2 samples are human; 0 samples in the whole SuperSeries carry KMT2A::AFF1. The human arm is K562, a BCR::ABL1 CML line with no KMT2A rearrangement, and supports only the H3K4me3 and Pol II conclusions.
counts
h3k79me2_chipseq_samples_…0.000h3k79me2_chipseq_samples_human: 0crispr_chip_h3k79me2_samp…0.000crispr_chip_h3k79me2_samples_human: 0rnaseq_samples_human0.000rnaseq_samples_human: 0samples_carrying_kmt2a_af…0.000samples_carrying_kmt2a_aff1: 0crispr_chip_h3k79me2_samp…2.000crispr_chip_h3k79me2_samples_total: 2h3k79me2_chipseq_samples_…10.0h3k79me2_chipseq_samples_total: 10rnaseq_samples_total24.0rnaseq_samples_total: 24
inputs
[
  "GSE192562",
  "GSE192544",
  "GSE192555",
  "GSE192561",
  "PMID 37679565",
  "bioRxiv 10.1101/2023.03.17.533231",
  "UniProt P51825, P42568, Q03111"
]
counts — the numbers behind the chart
{
  "h3k79me2_chipseq_samples_total": 10,
  "h3k79me2_chipseq_samples_human": 0,
  "crispr_chip_h3k79me2_samples_total": 2,
  "crispr_chip_h3k79me2_samples_human": 0,
  "rnaseq_samples_total": 24,
  "rnaseq_samples_human": 0,
  "samples_carrying_kmt2a_aff1": 0
}
How this could bite us back
  • If a pharmacodynamic endpoint for menin inhibitors in infant KMT2A::AFF1 disease is designed from this partition, it is being designed from a mouse myeloid measurement in a different fusion, and a null result in patients becomes uninterpretable - it could mean the drug failed engagement or that the partition does not exist in that genotype.
  • The correct response is to repeat the experiment, not to discard it: a project that reacts to a provenance failure by deleting the evidence loses the only unbiased functional-genomic support the menin interface has.
  • Residual H3K79me2 after menin inhibition was predicted to be MLLT10-dependent; if in KMT2A::AFF1 cells the residue instead reflects a route the fusion itself uses, a combination designed to remove 'the endogenous unit' would be removing part of the oncogenic one and losing the selectivity that made the menin inhibitor tolerable.
Still open
  1. Does the menin-dependent / MLLT10-directed partition of H3K79 methylation exist in a human KMT2A::AFF1 line such as SEM or RS4;11, where the fusion carries no AHD of its own?
  2. Which menin inhibitor, at what exposure, produced the residual H3K79me2 in GSM5751590? The GEO metadata does not name it and the full text is unreachable, so the degree of target engagement behind that key result is unknown to this graph.
  3. Is the mouse MLL-AF9 line here a freshly transduced primary line or a serially transplanted one? Transplantation selects, and a partition measured in a selected line need not be the partition present at diagnosis.
Every source behind this entry (6)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    The published CRISPR-ChIP paper reports a functional partitioning of H3K79 methylation into an oncogenic DOT1L complex directed by the MLL fusion in a menin-dependent manner and a separate endogenous DOT1L complex directed by MLLT10, and neither its title nor its abstract names the MLL translocation used.

    CRISPR-ChIP reveals selective regulation of H3K79me2 by Menin in MLL leukemia. · Gilan O, Talarmain L, Bell CC, Neville D, Knezevic K, Ferguson DT, Boudes M, Chan YC, Davidovich C, Lam EYN, Dawson MA · Nat Struct Mol Biol 2023
    PMID 37679565DOI
    What this evidence cannot show

    Abstract read from the NCBI efetch record on 2026-09-19; the full text could not be read. Routes tried and failed: Europe PMC (isOpenAccess N, no PMCID, inEPMC N); Unpaywall (oa_status 'closed', best_oa_location null); Semantic Scholar (openAccessPdf empty, status CLOSED); and the bioRxiv preprint PDF, .full HTML and JATS source XML, each of which returned HTTP 429 with Cloudflare error 1015 across four attempts and two clients. This entry therefore asserts nothing about the paper's figures or Methods text - only about its deposited data.

    Evidence that argues against this (1)
    • A GEO annotation is submitter-entered metadata, not a Methods section. If the authors ran further H3K79 experiments in a human MLL-AF4 line without depositing sequencing - a western blot or a calibrated ChIP-qPCR - those would not appear in GEO and this entry would understate the study's scope. The bound is on deposited sequencing, not on every experiment performed. Gilan O et al. 2023
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset

    GEO series GSE192544 (CRISPR-ChIP) contains four human K562 (CML) samples supporting the H3K4me3 and Pol II arms and two Mus musculus samples annotated 'source: AML, cell type: MLL-AF9' (GSM5751161 'MLLAF9 screen control'; GSM5751162 'MLLAF9 H3K79me2 CRISPR-ChIP day 9'), and its shared treatment protocol names only K562 and MLLAF9 Cas9 cells.

    CRISPR-ChIP to identify mechanisms of chromatin regulation [CRISPR-ChIP] (GSE192544) · Gilan O, Dawson MA, et al. (GEO submitters) · NCBI Gene Expression Omnibus 2023
    What this evidence cannot show

    Read on 2026-09-19 from the GEO accession-display text endpoint (targ=all&form=text&view=brief) and counted programmatically. Submitter-entered fields. The mouse MLL-AF9 line is not further specified - no indication of whether it is a freshly transduced primary line, a serially transplanted line, or an established cell line - and the sgRNA library is described only as a custom targeted epigenetic library, so the screen's coverage of the genome is not bounded by this record.

    Evidence that argues against this (1)
    • The SuperSeries is annotated 'Homo sapiens; Mus musculus', which invites the reading that humans feature in every arm. They do not: the mixed taxon reflects the human K562 H3K4me3/Pol II arm plus a deliberate mouse/human library-mix control ('Mouse/human library mix DMSO K562 cells'), not a human H3K79me2 experiment. Gilan O et al. 2023
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset

    In GEO series GSE192555 every H3K79me2 ChIP-seq sample is Mus musculus, cell type MLLAF9 - GSM5751582 control guide, GSM5751583-4 MLLT10 KO guides 1 and 2, GSM5751585-6 DOT1L KO guides 1 and 2, GSM5751589 DMSO, GSM5751590 menin-inhibitor treated, GSM5751591 DOT1L-inhibitor treated - together with GSM5751587-8, the DMSO and menin-inhibitor-treated FLAG-MLLAF9 occupancy samples; the thirteen human samples in the same series are K562 and carry only Pol II, H3K4me3 and H3K27ac.

    CRISPR-ChIP to identify mechanisms of chromatin regulation [ChIP-seq] (GSE192555) · Gilan O, Dawson MA, et al. (GEO submitters) · NCBI Gene Expression Omnibus 2023
    What this evidence cannot show

    Read on 2026-09-19 from the GEO accession-display text endpoint and enumerated sample by sample. Four human samples carry the evidently typographic cell-type values 'K563', 'K564', 'K565' and 'K566', which we read as K562; this touches only the human arm and none of the H3K79me2 samples. The menin inhibitor is not named anywhere in the GEO metadata, so its identity, dose and exposure are unknown to us.

    Evidence that argues against this (1)
    • This is also the evidence that makes the partition credible: MLLT10 knockout and menin inhibition are separated experimentally on the same mark in the same cells, with both a genetic and a pharmacological arm. The provenance finding cuts both ways - the internal design is strong and it is only the external transfer to KMT2A::AFF1 that is unsupported. Gilan O et al. 2023
  4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset

    GEO series GSE192561 (RNA-seq) comprises 24 samples, all Mus musculus with source name AML, covering non-targeting control, DOT1L knockout, and two guides each against MLLT10, MLLT1/ENL and MLLT6/AF17, in triplicate.

    CRISPR-ChIP to identify mechanisms of chromatin regulation [RNA-seq] (GSE192561) · Gilan O, Dawson MA, et al. (GEO submitters) · NCBI Gene Expression Omnibus 2023
    What this evidence cannot show

    Read on 2026-09-19 from the GEO accession-display text endpoint. Sample titles are submitter shorthand and this subseries' per-sample characteristics do not restate the fusion; we infer the same MLL-AF9 line from the shared study design and the identical AML source-name annotation, which is an inference, not a stated fact.

    Evidence that argues against this (1)
    • The transcriptional arm knocks out MLLT10, MLLT1/ENL and MLLT6/AF17 - the three DotCom/AHD-family candidates - and does not test AFF1 or AFF4. An entry that reads the partition as a general statement about MLL fusions is reading past the genes actually targeted. Mohan M et al. 2010
  5. E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    The ANC1 homology domain (AHD) of AF9 is intrinsically disordered and recruits multiple partners - including the elongation factor AF4 and, per the same laboratory's follow-up, DOT1L - through coupled folding and binding, with conserved aliphatic residues from each partner forming an integral part of the complex's hydrophobic core.

    PMID 23260655DOI
    What this evidence cannot show

    Abstract fetched from PubMed and read on 2026-09-19; full text not read, every identifier copied from that fetched record. NMR of isolated domains and peptides, not of a fusion protein in chromatin. The structure solved in this paper is the AF9 AHD with AF4; the DOT1L complex is from the follow-up (PMID 25921540). 'The same hydrophobic core' is our compression of the abstract's statement about conserved aliphatic residues, which implies a shared site without by itself proving one.

    Evidence that argues against this (1)
    • DOT1L has three separate AF9-binding sites and structure-guided point mutations produce graded rather than all-or-none loss of recruitment, so a single-site, all-or-nothing competition model oversimplifies the AHD interface. Kuntimaddi A et al. 2015
  6. E6established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProtKB P51825 (primary gene name AFF1, 1,210 aa) cross-references exactly three Pfam families - PF05110 (AF-4), PF18875 (AF4_int) and PF18876 (AFF4_CHD) - and no AHD family, whereas P42568 (MLLT3, 568 aa) and Q03111 (MLLT1, 559 aa) each carry PF17793 (AHD) alongside PF03366 (YEATS).

    What this evidence cannot show

    Each accession was fetched directly by accession on 2026-09-19 and its PRIMARY gene name and sequence length read back, not resolved through a gene_exact query - the failure mode that turned TCF3 into TCF7L1 in cycle 20. Pfam coverage is model-based annotation: absence of an assignment is absence of a detected match, which for a protein UniProt annotates as overwhelmingly disordered is weaker evidence than it would be for a globular protein.

    Evidence that argues against this (1)
    • Domain absence is not functional absence. AFF1 carries PF18875, the AF4 interaction motif, which is precisely the element that binds the AF9/ENL AHD - so the index fusion can reach an AHD in trans although it contains none. Whether that route delivers DOT1L is a separate question, and reconstitution says it does not. Biswas D et al. 2011
Connected to