Oncofusion

start / graph / ana-q147-under-revumenib-hoxa-occupancy-collapses-and-the-irx1-question-becomes-unanswerable

Analysismoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q147-hoxa-irx1-across-partners

Under revumenib the oncoprotein leaves chromatin globally, so 'does IRX1 become a menin-inhibitor escape route' cannot be answered from this deposit

In the one child in this dataset sampled while on a menin-blocking drug, the fusion protein has come off the HOXA genes - and off everything else too. So you cannot use that sample to ask whether it moved to IRX1 instead.

The same thing, for a scientist

GSE252378 contains one revumenib-exposed aliquot, **152985** (KMT2A-ENL, infant AML at relapse), paired with the drug-naive diagnostic aliquot **148752** (infant B-ALL) from the same child. | | 148752 (naive) | 152985 (on revumenib) | |---|---:|---:| | p_dz (depth) | 0.0000 | 0.0000 | | HOXA9-HOXA10 | **9.69x** | **0.97x** | | MEIS1 (2 kb, N1/N2) | 9.2 / 13.3x | **1.6 / 0.8x** | | CDK6 (2 kb, N1/N2) | 8.4 / 15.7x | **1.2 / 1.7x** | | IRX1 gene body | 0.25x | 0.68x | | own floor | 0.31x | **1.03x** | | readability gate | pass | **FAIL** | **152985 fails the pre-registered readability gate**: neither MEIS1 nor CDK6 clears 2.0x in either KMT2A-N library, at the *best sequencing depth in the deposit* (0.1-0.3% of promoters zero) [E1]. That is not a bad library - it is the expected consequence of menin inhibition, and it is the deposit paper's own finding that in this sample the oncoprotein and menin fall to undetectable levels while ENL persists at target loci [E2]. **The consequence, and it is the point of this entry:** IRX1 reads 0.68x in 152985, which is **below its own floor of 1.03x** and therefore not occupied - but because no locus in that library is occupied, **the sample cannot distinguish 'IRX1 is not a menin-inhibitor escape route' from 'this library cannot report any route'** [E1]. A reader who saw only the IRX1 number would conclude the former. The pre-registered readability gate is the only thing that stops that conclusion being drawn, and it must not be drawn.

Go deeper
Why the floor rises under the drug, and why that is informative

152985's own cannot-be-occupied floor is 1.03x, against a deposit-wide median of 0.22x and a maximum elsewhere of 1.59x. Under global eviction the within-sample share flattens: when nothing is enriched, every locus tends towards 1.0x by construction. That is the signature of the state, and it is why a within-sample share is the wrong instrument for measuring the drug effect and the right instrument for detecting that the drug worked.

This is the same shape as the corpus rule that a share says relative to average, not gained. Here the flattening is the result.

The experiment that would answer the question

To ask whether IRX1 becomes an escape address under menin inhibition you need a sample in which the oncoprotein is still on chromatin somewhere - i.e. partial rather than complete eviction - or a spike-in-calibrated absolute measurement. Neither exists in GSE252378: there is one drug-exposed aliquot, it is completely evicted, and there is no spike-in.

And the escape question is specifically about KMT2A::AFF1, because that is the only partner in which IRX1 is ever fusion-occupied. There is no revumenib-exposed KMT2A::AFF1 aliquot in this deposit at all. See exp-q147-irx1-occupancy-under-partial-menin-inhibition-in-kmt2a-aff1.

Facts and figures
drugrevumenib
drug exposed aliquot152985
drug naive paired aliquot148752
partnerKMT2A-ENL
hoxa naive9.689
hoxa treated0.97
irx1 treated0.684
floor treated1.03
readability treatedFAIL
p dz both0
verdictthe IRX1 escape question is UNANSWERABLE from this deposit
meis1 naive
[
  9.22,
  13.31
]
meis1 treated
[
  1.56,
  0.82
]
cdk6 naive
[
  8.42,
  15.65
]
cdk6 treated
[
  1.18,
  1.73
]
analyses
[
  "ana-q147-irx1-fusion-occupancy-is-confined-to-kmt2a-aff1"
]
How this could bite us back
  • This is the bite-back for the menin-interface hypothesis stated as a measurement gap rather than a speculation: the one place the graph could have looked for a post-menin-inhibitor escape address, it cannot look, because complete eviction destroys the instrument. Any claim that IRX1 is or is not an escape route is currently unevidenced in either direction.
  • A within-sample share flattens towards 1.0 under global eviction, so a drug that works will make every locus look 'average' and a naive reader will see rising shares at previously low loci. That is an artefact of renormalisation and it would look exactly like an escape route.
Still open
  1. Is there any deposited KMT2A::AFF1 sample under a menin inhibitor with partial rather than complete oncoprotein eviction? That is the only material in which the IRX1 escape question is askable, and GSE252378 has none.
  2. Would a spike-in-calibrated CUT&RUN under revumenib show IRX1 occupancy rising in absolute terms while every within-sample share flattens? The two would look identical without the spike-in.
Every source behind this entry (2)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    In GSE252378, the revumenib-exposed aliquot 152985 has KMT2A-N shares of 0.97x at HOXA9-HOXA10, 1.6x/0.8x at MEIS1 and 1.2x/1.7x at CDK6, failing the pre-registered readability gate, while its paired drug-naive aliquot 148752 reads 9.69x, 9.2x/13.3x and 8.4x/15.7x respectively. Both have p_dz = 0.0000, the best depth in the deposit. IRX1 in 152985 is 0.68x against that library's own cannot-be-occupied floor of 1.03x - the highest floor of any KMT2A-N library in the deposit.

    What this evidence cannot show

    n = 1 drug-exposed aliquot, and it is confounded with relapse, with a lineage switch from B-ALL to AML, and with clonal replacement - this corpus already records that drug exposure and clonal replacement are completely confounded across this pair. No dose or exposure duration is recorded anywhere in the deposit. Nothing here is a drug-effect measurement; it is a statement about what the sample can and cannot be asked. Every number is a WITHIN-SAMPLE share: an interval's mean coverage divided by that same library's mean coverage over all 19,404 hg38 RefSeq promoters, so 1.0 = an average promoter in that library. GSE252378 has no spike-in, so no between-library absolute statement is licensed; the only cross-sample quantity used is a ratio of two loci measured inside one library, where the unknown per-library scale cancels. The occupancy call is absolute, not rank-based: >=2.0x an average promoter in BOTH KMT2A-N antibodies AND above that library's own floor, where the floor is the 95th percentile of eight width-matched windows at loci that cannot be occupied in a leukaemia (ALB, RHO, MYH7, MYOD1, NEUROD1, KRT14, GFAP, INS). IgG is not the gate: q145 measured GSE252378 IgG libraries as 57-94% zero at promoters, and one of them (TB051939) reads 13.1x an average promoter at a locus that cannot be occupied.

    Evidence that argues against this (1)
    • The pair is one patient at two disease states, so the difference between 9.69x and 0.97x at HOXA is not attributable to the drug: the corpus entry `ana-q143-drug-exposure-and-clonal-replacement-are-completely-confounded-across-this-patient-pair` establishes that the two cannot be separated in this material. Iqbal Lab oncofusion knowledge graph et al. 2026
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The deposit's own paper reports that in the sample that lineage-switched during treatment with revumenib, the oncoprotein and menin are reduced to undetectable levels, while ENL - a transcriptional cofactor of the oncoprotein - persists on numerous oncoprotein-target loci including genes in the GMP-like lineage-switching programme.

    MLL oncoprotein levels influence leukemia lineage identities · Janssens DH, Duran M, Otto DJ, Wu W, Xu Y, Kirkey D, Mullighan CG, Yi JS, Meshinchi S, Sarthy JF, Ahmad K, Henikoff S · Nat Commun 15:9341 2024
    PMID 39472576DOI
    What this evidence cannot show

    The paper's own interpretation of the same sample, which independently supports reading 152985's failure of the positive-control gate as biology rather than as a technical failure. It also warns against the converse error: ENL persistence means a cofactor signal at a locus does not establish that the oncoprotein is there.

    Evidence that argues against this (1)
    • If ENL persists at target loci without the oncoprotein, then any escape-route argument built on cofactor occupancy rather than on KMT2A-N occupancy would find activity that the oncoprotein is not responsible for - which cuts against using cofactor tracks as a workaround for the unreadable KMT2A-N libraries. Janssens DH et al. 2024
Connected to