Oncofusion

start / graph / exp-q147-kmt2a-n-cut-and-run-in-an-untreated-non-aff1-lineage-switch-blast

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

KMT2A-N CUT&RUN in an untreated non-AFF1 lineage-switching blast: the one experiment that separates partner from cell state

All three leukaemias where the fusion sits on IRX1 are the same fusion type AND the same kind of sample from the same hospital. One untreated sample of a different fusion type but the same kind would tell you which of those two things matters.

The same thing, for a scientist

**The confound this experiment exists to break.** The three IRX1-occupied samples in GSE252378 (p279, p318, p454) are all KMT2A::AFF1, **all from Texas Children's Hospital**, and all lineage-switch-associated infant cases. Partner, institution and cell state are completely confounded, so the deposit cannot say which produces the IRX1 address [E1]. The deposit contains the exact matched design - an **untreated** non-AFF1 infant lineage-switching pair from Texas Children's - and it is unusable, because the one ENL lineage-switch pair (148752/152985) has its AML aliquot on revumenib with the oncoprotein completely evicted [E1]. **Material.** An untreated infant KMT2A-ENL, -AF9, -AFDN or -AF10 lineage-switching case, sampled at both the lymphoid and the myeloid state, drug-naive at both. One case is informative; three would settle it. Primary blasts or PDX. **Tracks, per aliquot.** KMT2A-N1 (Millipore 05-764) and KMT2A-N2 (CST D2M7U) carried separately - they are not replicates; KMT2A-C1/C2 for the N-over-C contrast; menin; H3K4me3; ENL. Sequence to a promoter zero-rate below 20% per N library, so that p_dz lands under 0.05 and a zero at IRX1 is a measurement. **Primary readout.** KMT2A-N share at the IRX1 gene body (chr5:3,595,831-3,601,403, hg38) and at the HOXA9-HOXA10 span (chr7:27,159,437-27,177,320), each normalised within sample. **The pre-declared decision rule, which is the point of specifying it in advance:** - IRX1 >= 2.0x in both N antibodies and above the library floor, in the HOXA-low aliquot => **the IRX1 address is a cell-state property and this beat's partner reading is wrong**. - IRX1 below floor in both aliquots while HOXA is >= 2.0x in at least one => **the partner reading survives a matched cell state**, which is the strongest form of this beat's claim it could have. - HOXA below floor in both aliquots and MEIS1/CDK6 also below floor => the sample is globally low-oncoprotein and **uninformative**; say so and do not score IRX1 from it.

Go deeper
Why this is cheaper than it looks

AutoCUT&RUN on a primary aliquot needs of order 10^5 nuclei and the antibody panel is commercially available and already benchmarked in this exact assay by the deposit's authors. The scarce resource is the material: a drug-naive infant KMT2A-r lineage-switching pair. Texas Children's, St. Jude, Fred Hutch and Seattle Children's have all contributed such cases to GSE252378, so the biobanks exist.

Pre-declare the controls this beat used and do not substitute IgG for them: MEIS1 and CDK6 as within-sample positives in KMT2A-N (PAX5 only in B-lineage material); the eight-locus width-matched cannot-be-occupied panel (ALB, RHO, MYH7, MYOD1, NEUROD1, KRT14, GFAP, INS) as the per-library floor; and a depth report as the promoter double-zero probability p_dz with a stated cut. GSE252378's IgG libraries are 57-94% zero at promoters and reach 13.1x an average promoter at a locus that cannot be occupied, so IgG alone will not do.

Facts and figures
designpaired lymphoid and myeloid aliquots from one untreated non-AFF1 infant KMT2A-r lineage-switching case
assayAutoCUT&RUN or CUT&RUN, hg38
depth requirementpromoter zero-rate < 20% per KMT2A-N library (p_dz < 0.05)
min n1
preferred n3
decision ruleIRX1 >= 2.0x in both N antibodies and above the library floor in the HOXA-low aliquot refutes the partner reading
tracks
[
  "KMT2A-N1 Millipore 05-764",
  "KMT2A-N2 CST D2M7U",
  "KMT2A-C1",
  "KMT2A-C2",
  "MEN1",
  "H3K4me3",
  "ENL",
  "IgG"
]
primary intervals
{
  "IRX1_gene": "chr5:3,595,831-3,601,403",
  "HOXA9_HOXA10": "chr7:27,159,437-27,177,320"
}
tests hypotheses
[
  "hyp-q147-the-irx1-address-requires-the-aff1-partner-not-the-menin-interface",
  "hyp-q147-irx1-occupancy-is-a-lineage-switch-state-confounded-with-one-institution"
]
analyses
[
  "ana-q147-irx1-fusion-occupancy-is-confined-to-kmt2a-aff1"
]
How this could bite us back
  • If the result is that IRX1 occupancy IS a cell-state property, then it is reachable in every KMT2A partner and a menin inhibitor could drive patients into it - which makes the negative result of this beat the reassuring one and the positive result the clinically alarming one. The experiment should be powered and interpreted with that asymmetry in mind.
Still open
  1. Does an untreated non-AFF1 lineage-switching pair exist in any biobank? All four contributing institutions have supplied lineage-switch cases, so the question is whether a drug-naive myeloid aliquot was banked.
Every source behind this entry (1)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    All three IRX1-occupied samples in GSE252378 are KMT2A::AFF1 from Texas Children's Hospital and lineage-switch-associated; the deposit's only non-AFF1 infant lineage-switching pair from the same institution (148752/152985, KMT2A-ENL) has its AML aliquot on revumenib with MEIS1 at 1.6x/0.8x and CDK6 at 1.2x/1.7x, i.e. the oncoprotein off chromatin, so it cannot serve as the matched comparison.

    What this evidence cannot show

    The confound is structural in the deposit and is stated as the reason for the experiment, not as a result. Whether institution matters at all is unknown - it is a proxy for referral pattern, collection protocol and patient population, any of which could be the operative variable, or none. 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.

    Evidence that argues against this (1)
    • The deposit paper attributes the distinctive behaviour of lineage-switching samples to reduced oncoprotein LEVELS rather than to partner identity, which is the cell-state reading this experiment is designed to test and which is currently the better-supported of the two. Janssens DH et al. 2024
Connected to