Oncofusion

start / graph / hyp-menin-interface-evidence-is-partner-imported

Hypothesisspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.Q-partner-provenance

What could refute the menin-interface bet, restated after the provenance audit: three tests, none of which is about menin binding

also called menin interface falsifiers cycle 49 · is the menin evidence AF9-shaped

This project bets that the sticky point between menin and the fusion protein is the thing to attack. That bet is safe in one way - every version of the fusion keeps the menin-binding piece - and exposed in another, because the evidence that attacking it removes the disease's chromatin programme came from a different fusion in mice.

The same thing, for a scientist

**Statement.** The menin-interface hypothesis is unusually robust to partner import at its *binding* step and unusually exposed at its *consequence* step, and the audit shows the graph has not distinguished the two. **Robust:** menin binds the KMT2A N-terminus, which every translocation retains by definition, so the binding claim is partner-independent by construction. Nothing in this audit touches it. **Exposed:** the claim that menin inhibition works *because* it displaces the fusion and thereby collapses a fusion-directed chromatin programme rests on (i) the CRISPR-ChIP partition, now shown to be murine MLL-AF9 [E1]; (ii) the wild-type-allele requirement, murine MLL-AF9 [E2]; and (iii) the observation that menin spreads with MLL-AF4 into gene bodies alongside H3K79me2/3, which is index-fusion and correlative [E3]. **And there is a third, older exposure the audit sharpens:** menin dependency is not confined to KMT2A fusions. NUP98-rearranged leukaemia is menin-dependent [E4]. A dependency that appears across fusion families is a dependency on wild-type KMT2A-menin, a complex normal haematopoiesis also uses - which is an argument about **selectivity**, not efficacy, and it is not a partner-import artefact. Prior probability that the *consequence* account survives an index-fusion test as currently stated: 0.45.

Go deeper
Three tests, in order of how much they would cost the project

Test 1 (cheap, decisive for the chromatin narrative). Menin inhibition at verified full engagement in a human KMT2A::AFF1 line, calibrated H3K79me2 readout, with a KMT2A::MLLT3 line alongside. If H3K79me2 barely moves in the AFF1 line while dropping in the MLLT3 line, the chromatin account of menin inhibition is AF9-specific. The drug still works; the story does not. exp-crispr-chip-h3k79-partition-in-human-kmt2a-aff1.

Test 2 (moderate, decisive for target selection). Whether menin-inhibitor response in patients differs by 3' partner at a resolution finer than KMT2A-rearranged versus not. If KMT2A::MLLT10 - the genotype where the two H3K79 units converge - responds differently, the partition is real and partner-dependent. If nothing differs by partner, the mechanism may not run through the partner at all, and the whole import question is less consequential than this beat argues.

Test 3 (already partly answered, and the worst for selectivity). How many non-KMT2A leukaemias are menin-dependent. Each additional one strengthens efficacy and weakens the claim that the interface is a fusion-specific surface 4.

What this audit does NOT threaten, stated so the finding is not over-read
  • That menin binds the fusion. Structural and biochemical; partner-independent.
  • That menin inhibitors have clinical activity in KMT2A-rearranged leukaemia. Clinical, and not derived from any of the imported mechanism.
  • That MEN1 is a strong dependency in KMT2A-rearranged lines. A DepMap fact, measured in human lines including index-fusion ones.

What it threatens is the middle of the chain: menin displacement to fusion eviction to loss of fusion-directed H3K79 methylation to programme collapse. The third link in that chain is where the AF9 import sits, and where the index fusion has no mapped interface (mechanism-kmt2a-aff1-dot1l-recruitment-route).

The honest version of the beat's headline

Six beats of counter-argument work in this graph have looked for evidence against the menin interface and found mostly selectivity arguments. The provenance audit adds a different kind of challenge: not is the claim false but was it ever measured here. For the binding step the answer is yes. For the chromatin-consequence step the answer is no, and the paper that looked most like a yes turned out to be a mouse MLL-AF9 experiment 1.

That is a weaker attack than a contradicting result and a stronger one than a caveat, and it is repairable by a single experiment - which is the best kind of problem to have.

Facts and figures
statementThe menin-interface hypothesis is partner-independent at its binding step and partner-imported at its chromatin-consequence step; the graph has not distinguished the two, and the distinction determines which falsifiers matter.
origindata-pattern
prior probability0.45
predictions
[
  "Menin inhibition reduces H3K79me2 at fusion targets substantially less in a human KMT2A::AFF1 line than in a matched KMT2A::MLLT3 line.",
  "Menin-inhibitor clinical response does not differ by 3' partner except in KMT2A::MLLT10, where the two H3K79 units converge.",
  "Further non-KMT2A menin-dependent leukaemias will be reported, strengthening efficacy and weakening fusion-specific selectivity."
]
experiments
[
  "exp-crispr-chip-h3k79-partition-in-human-kmt2a-aff1",
  "exp-ahd-occupancy-competition-in-t411-cells"
]
analyses
[
  "ana-partner-provenance-audit",
  "ana-crispr-chip-fusion-identity-resolved"
]
falsifiers
[
  "Menin inhibition removes H3K79me2 at fusion targets in a human KMT2A::AFF1 line as efficiently as in MLL-AF9 - the chromatin-consequence step would then be partner-independent after all and this hypothesis is wrong.",
  "A residue-level KMT2A::AFF1-DOT1L contact is mapped - the index fusion would then have a fusion-directed unit and the import would have been harmless.",
  "Menin-inhibitor response differs strongly by 3' partner in a way that tracks the presence of a DOT1L interface - which would confirm the partition and make the import a real clinical variable rather than a bookkeeping one.",
  "No further non-KMT2A leukaemia proves menin-dependent on adequate testing - the selectivity argument would then weaken on its own."
]
How this could bite us back
  • If the chromatin-consequence account is AF9-specific, biomarker development for menin inhibitors in infant disease is currently pointed at the wrong analyte, and a trial could conclude that a working drug failed to engage.
  • Menin dependency in non-KMT2A leukaemias widens the indication and narrows the safety argument at the same time: a target normal haematopoiesis uses, hit in more diseases, for longer.
  • The audit's finding is repairable by one experiment, which makes it tempting to defer; deferring it keeps the graph's central chain resting on a mouse myeloid measurement in a different fusion.
Still open
  1. Have the menin-inhibitor trials reported outcome by 3' partner? This graph does not hold that breakdown, and it is the cheapest of the three tests because the data already exist.
  2. Is there any experiment in which menin inhibition was applied to a human KMT2A::AFF1 line with a calibrated H3K79 readout? This beat found none.
  3. If menin inhibition acts substantially through wild-type KMT2A rather than the fusion, does the import question matter for menin at all - or only for DOT1L?
Every source behind this entry (5)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset

    Every H3K79me2 experiment in the CRISPR-ChIP study, including the menin-inhibitor-treated arm, is annotated in GEO as Mus musculus, cell type MLLAF9, source AML; no human sample in the study carries a KMT2A rearrangement.

    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. Submitter-entered metadata; bounds the deposited sequencing rather than every experiment performed. The menin inhibitor is unnamed in the record, so target engagement in the source experiment is unknown to this graph.

    Evidence that argues against this (1)
    • A mouse MLL-AF9 result is not a wrong result, and the partition is the most assumption-free support the menin interface has received. Treating an import as a refutation would be the mirror error of treating it as a demonstration. Gilan O et al. 2023
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The wild-type Mll allele is required for MLL-AF9-induced leukemogenesis and for maintenance of transformed cells, with menin recruiting both wild-type MLL and the fusion to HOX loci and wild-type MLL controlling a distinct component of the programme.

    MLL-AF9-induced leukemogenesis requires coexpression of the wild-type Mll allele. · Thiel AT, Blessington P, Zou T, Feather D, Wu X, Yan J, Zhang H, Liu Z, Ernst P, Koretzky GA, Hua X · Cancer Cell 2010
    PMID 20159607DOI
    What this evidence cannot show

    Abstract fetched from PubMed and read on 2026-09-19; full text not read. Mouse genetics in one fusion, transplantation model, binary 'required' design; no human allele-dosage data exist and the experiment has not been repeated in a KMT2A::AFF1 background.

    Evidence that argues against this (1)
    • This result is also the strongest argument that menin inhibition acts partly through wild-type KMT2A rather than through the fusion, which would make the drug's action partner-independent - and would mean the import question matters less for menin than for DOT1L. Heikamp EB et al. 2022
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    MLL-AF4 binding spreads into gene bodies at a subset of targets, accompanied by spreading of menin binding, increased transcription and increased H3K79me2/3.

    MLL-AF4 Spreading Identifies Binding Sites that Are Distinct from Super-Enhancers and that Govern Sensitivity to DOT1L Inhibition in Leukemia. · Kerry J, Godfrey L, Repapi E, Tapia M, Blackledge NP, Ma H, Ballabio E, O'Byrne S, Ponthan F, Heidenreich O, Roy A, Roberts I, Konopleva M, Klose RJ, Geng H, Milne TA · Cell Rep 2017
    PMID 28076791DOI
    What this evidence cannot show

    Abstract fetched from PubMed and read on 2026-09-19; full text not read. Human MLL-AF4 ChIP-seq with lines unnamed in the abstract. Correlative co-occupancy: menin, the fusion and the mark being in the same place does not establish that removing menin removes the mark.

    Evidence that argues against this (1)
    • Menin spreading, fusion spreading and H3K79me2/3 could all follow elevated transcription, in which case the co-occupancy is a consequence of output rather than evidence for a menin-to-H3K79 causal chain. Godfrey L et al. 2019
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The menin-MLL1 interaction is a molecular dependency in NUP98-rearranged leukaemia, a disease not driven by a KMT2A fusion.

    The menin-MLL1 interaction is a molecular dependency in NUP98-rearranged AML. · Heikamp EB, Henrich JA, Perner F, Wong EM, Hatton C, Wen Y, Barwe SP, Gopalakrishnapillai A, Xu H, Uckelmann HJ, Takao S, Kazansky Y, Pikman Y, McGeehan GM, Kolb EA, Kentsis A, Armstrong SA · Blood 2022
    PMID 34582559DOI
    What this evidence cannot show

    Identifier, title, journal, year and DOI fetched from PubMed on 2026-09-19 and copied from that record; abstract not re-read this cycle, so the claim is asserted at the level of the title's statement. NUP98-rearranged disease is myeloid.

    Evidence that argues against this (1)
    • A broader menin dependency could reflect that NUP98 fusions co-opt wild-type KMT2A, in which case menin inhibition is still acting on a leukaemia-specific configuration rather than on generic menin function - so cross-family dependency is not automatically a selectivity problem. Thiel AT et al. 2010
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    In a stratified random sample of 48 mechanism-bearing entries in this graph, 13 of the 24 whose load-bearing evidence involves a fusion (54%) come from a fusion other than KMT2A::AFF1, and none rests on a dedicated human KMT2A::AFF1 experiment.

    What this evidence cannot show

    Our own count, 2026-09-19, classifying only the first evidence item of each sampled entry and stratifying by folder rather than proportionally; n = 48 gives the 54% figure a binomial 95% interval of roughly 33-74%. See ana-partner-provenance-audit for the full method and its three stated limitations.

    Evidence that argues against this (1)
    • Twelve of those 13 entries already flag their own import in caveats, so the graph's per-entry honesty is high; the audit's finding is about the missing aggregate, not about entries misrepresenting their sources. Oncofusion knowledge graph et al. 2026
Connected to