Oncofusion

start / graph / hyp-q143-the-menin-that-matters-in-an-infant-is-the-menin-on-the-fusion-and-the-n-minus-c-difference-can-measure-it-in-any-patient-sample

Hypothesisspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.Q143-menin-under-drug-in-infants

The menin that matters in an infant KMT2A-rearranged leukaemia is the menin on the fusion, and the N-minus-C antibody difference can measure it in any patient sample without a tag

We think the menin worth targeting is the menin attached to the broken fusion protein, not the menin attached to the normal version of the same gene. And we think there is a cheap way to tell them apart in real patient samples: compare an antibody that sees the front half of the protein with one that sees the back half.

The same thing, for a scientist

Two linked propositions. **(1) Compartment.** In primary infant KMT2A-rearranged leukaemia, menin's chromatin residency partitions into a fusion-associated pool at oncoprotein-target promoters and a wild-type-MLL1-associated pool elsewhere, and only the first is therapeutically relevant. **(2) Instrument.** Because taspase 1 cleaves wild-type KMT2A into N320 and C180 fragments that stay associated while a fusion retains only the N-terminal portion, the **difference between an N-terminal and a C-terminal antibody's occupancy** identifies fusion-occupied sites in unmodified patient material. The first observation supporting both is in this beat: in a drug-naive infant KMT2A::MLLT1 sample, the N channel recovers 40 of the top 50 published oncoprotein-target genes and the C channel 1 of 50, and menin's enrichment tracks the N channel (2.73x) and not the C channel (1.10x). If the instrument holds, every deposited primary KMT2A-r sample with paired N and C antibodies becomes a fusion-occupancy dataset retrospectively - and GSE252378 alone contains dozens.

Go deeper
Why it is worth stating as a hypothesis rather than a result

The supporting observation is n = 1 patient in one fusion, with an accessibility reference borrowed from a different child, and the C-channel null admits an epitope-masking explanation that this beat cannot exclude. What makes it a hypothesis worth testing rather than a curiosity is its leverage: if true, it converts an existing public deposit into a fusion-occupancy resource for primary human material, which this project has repeatedly recorded as the thing it most lacks.

Predictions, each scoreable
  1. In the untreated KMT2A::AFF1 samples of GSE252378 (p279, p454, SEM), the N channel will show >= 10-fold hypergeometric enrichment for the authors' oncoprotein-target gene set at k = 50 and the C channel will show < 2-fold. (Directly computable from public files.)
  2. Across those samples, menin's accessibility-corrected enrichment at the top 50 N-ranked promoters will exceed its enrichment at the top 50 C-ranked promoters by at least 1.5-fold.
  3. In a sample with a documented expressed reciprocal partner::KMT2A-C product, the C channel will show partial enrichment at oncoprotein targets, and the N-minus-C instrument will degrade accordingly.
  4. In an HA-tagged KMT2A::MLLT3 degron line, the N-minus-C difference computed from untagged antibodies will correlate with the HA channel at Spearman >= 0.6 over promoters. (This is the calibration that would turn the instrument quantitative.)
Falsifiers
  • F1. If in three or more untreated primary KMT2A-r samples the C channel enriches for the authors' oncoprotein-target gene set at k = 50 as strongly as the N channel does (fold within 2x of each other, both p < 1e-5), the N/C difference is not fusion-specific and proposition 2 is dead. Scoreable on public files, no wet work.
  • F2. If in an HA-tagged degron line the N-minus-C difference correlates with the HA channel at Spearman < 0.3 over promoters, the instrument does not measure fusion occupancy. One cell line, existing antibodies.
  • F3. If a spike-in-normalised menin CUT&RUN in a KMT2A-r infant sample shows menin enrichment at wild-type-MLL1-only promoters (high C, low N) equal to or greater than at fusion promoters, proposition 1 is dead.
Facts and figures
statementmenin's therapeutically relevant chromatin pool in infant KMT2A-r leukaemia is the fusion-associated pool, and the N-terminal-minus-C-terminal antibody difference measures fusion occupancy in unmodified patient material
origindata-pattern
prior probability0.55
shown for fusionKMT2A::MLLT1 in one primary infant patient; untested in KMT2A::AFF1
predictions
[
  "in untreated KMT2A::AFF1 samples of GSE252378 the N channel enriches >=10-fold for the published oncoprotein-target set at k=50 and the C channel <2-fold",
  "menin's accessibility-corrected enrichment at N-ranked promoters exceeds that at C-ranked promoters by >=1.5-fold",
  "a sample expressing the reciprocal partner::KMT2A-C product degrades the instrument",
  "in an HA-tagged degron line the N-minus-C difference correlates with HA at Spearman >=0.6"
]
analyses
[
  "ana-q143-a-filter-free-promoter-ranking-recovers-forty-of-the-fifty-published-oncoprotein-target-genes",
  "ana-q143-menin-sits-with-the-fusion-n-terminus-and-not-with-the-wild-type-c-terminus-in-a-drug-naive-infant-b-all"
]
experiments
[
  "exp-q143-spike-in-normalised-menin-cut-and-run-with-paired-n-and-c-antibodies-in-an-infant-kmt2a-aff1-sample-on-and-off-drug"
]
falsifiers
[
  "F1: if in three or more untreated primary KMT2A-r samples the C channel enriches for the published oncoprotein-target gene set at k=50 within 2x of the N channel (both p<1e-5), the N/C difference is not fusion-specific. Scoreable on public files, no wet work.",
  "F2: if in an HA-tagged KMT2A::MLLT3 degron line the N-minus-C promoter difference correlates with the HA channel at Spearman <0.3, the instrument does not measure fusion occupancy.",
  "F3: if spike-in-normalised menin CUT&RUN in a KMT2A-r infant sample shows menin enrichment at wild-type-only promoters (high C, low N) at or above that at fusion promoters (high N, low C), the compartment proposition is dead."
]
How this could bite us back
  • If the N-minus-C readout becomes a pharmacodynamic assay, it reports epitope presence and not fusion activity - a fusion held inactive still carries its N-terminus.
  • A drug that degrades the fusion and a drug that displaces the whole complex would look different on this readout, which is useful; a drug that merely stops the fusion working would look like no effect, which is a false negative built into the assay.
Still open
  1. Can the N-minus-C instrument be calibrated against a tagged system so that it reports fusion occupancy quantitatively rather than as a ranking?
  2. How many public deposits carry paired N-terminal and C-terminal KMT2A antibodies on primary KMT2A-r material? GSE252378 does; a survey was not run.
  3. Does the instrument work for 3'-KMT2A fusions, where the fusion retains the C-terminus? For those the polarity inverts and the same logic applies in reverse.
Every source behind this entry (3)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    In a drug-naive infant KMT2A::MLLT1 sample the N-terminal channel recovers 40 of the top 50 published oncoprotein-target genes and the C-terminal channel 1 of 50, and menin's enrichment tracks the N channel.

    analyses/menin-under-revumenib-in-an-infant-patient-pair - pre-registered re-analysis of GSE252378 samples 148752 and 152985 · Q143-menin-under-drug-in-infants (Oncofusion Knowledge Graph) · this repository 2026
    What this evidence cannot show

    n = 1 patient, one aliquot, cross-assay comparison, out-of-sample accessibility reference, no spike-in, lineage-switching patient, and not the index fusion.

    Evidence that argues against this (1)
    • A C-terminal epitope can be masked at a busy promoter rather than absent, which would produce the same result without any fusion-specific interpretation. Olsen SN et al. 2022
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    Taspase 1 cleaves wild-type KMT2A into N320 and C180 fragments that re-associate, which is the biochemical basis for the instrument.

    What this evidence cannot show

    Accession and length fetched live and read back against the primary gene name KMT2A; the two processed-chain entries are curator statements without ECO:0000269 codes.

    Evidence that argues against this (1)
    • UniProt's two processed-chain location comments are uncoded curator statements rather than pointers to a specific localisation experiment, and this project has measured that UniProt's coded citations for KMT2A are largely not KMT2A localisation studies - so the database is a convenient statement of the biochemistry and not its founding source. UniProt Consortium 2026
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset

    GSE252378 runs paired N-terminal and C-terminal KMT2A antibodies on dozens of primary patient samples, which is the material the hypothesis would convert into a resource.

    What this evidence cannot show

    Counted from the antibody lines of the 171 sample records; for seven patients the raw sequencing is withheld and only aligned files are shared, which is sufficient for this analysis but not for genotyping.

    Evidence that argues against this (1)
    • For seven patients in this series the raw sequencing is withheld, so any retrospective resource built on the paired N and C antibodies cannot be accompanied by genotyping for those patients - which matters because the reciprocal product and the breakpoint both change how the N/C difference should be read. Janssens DH et al. 2024
Connected to