Oncofusion

start / graph / ana-q147-a73051-is-a-hoxa-negative-irx1-negative-kmt2a-af10-blast

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

A73051 is the one HOXA-negative non-AFF1 sample in GSE252378, and it is IRX1-negative too

One paediatric leukaemia in this dataset has almost no fusion protein on the HOXA genes - the same state that in t(4;11) goes with IRX1 being switched on. In this one, IRX1 is switched off as well.

The same thing, for a scientist

🟠 **UPDATE 2026-09-24 (cycle 135): this entry's own stated weakness has been tested and it cannot be resolved in this deposit.** The summary below says a globally low-oncoprotein sample *"cannot distinguish 'IRX1 not occupied' from 'nothing occupied'"*. q148 scored the authors' GMP-like class across the cohort and reported A73051 as **globally low**. On re-derivation that is **not established**: the oncoprotein share correlates with raw library coverage at **+0.659**, the floor-referenced ratio still at **+0.631**, and A73051's libraries are in the **bottom fifth** of coverage. With no spike-in, scarce protein *causes* shallow coverage, so the technical and biological readings are not separable. **The caveat stands, neither harder nor softer than this entry already stated it.** 🔵 **What is unaffected:** the 0-of-18 cohort bound does not rest on A73051 - the other 17 non-AFF1 samples are HOXA-occupied at **2.94-29.11x**, so they demonstrably carry oncoprotein on chromatin and none has it at IRX1. --- If the HOXA/IRX1 split were general to KMT2A-r, the informative sample is a HOXA-negative non-AFF1 blast. **GSE252378 contains exactly one: A73051, KMT2A-AF10, paediatric AML, Fred Hutch.** Its HOXA9-HOXA10 share is **0.870x** an average promoter (HOXA cluster 0.701x, HOXA9 promoter 0.709x, HOXA10 promoter 0.885x) - the lowest non-AFF1 HOXA in the deposit and comfortably NOT_OCCUPIED. Its IRX1 gene-body share is **0.249x against its own floor of 0.255x**, and its IRX1 promoter share is **exactly 0.000 in both N antibodies** [E1]. **It is depth-ADEQUATE** (p_dz = 0.0120, the fourth best in the deposit), so the zero is a measurement and not a gap [E1]. So the reciprocal phenotype does not appear where the model predicts it should. This is the single most informative non-AFF1 sample in the deposit for the question asked, and it points the same way as the aggregate. **The honest weakness, stated because it is the whole caveat:** A73051 has the weakest positive controls in the deposit. MEIS1 is 2.42x and 2.61x (clears the 2.0x gate by 21%), **CDK6 is 0.61x and 0.66x and fails it outright**, and at 8 kb MEIS1 falls to 1.26x. It passes the pre-registered readability gate on MEIS1 alone, which is exactly what the gate was written to allow, and it is entirely possible that this is a globally low-oncoprotein sample rather than a HOXA-selective one [E1]. A globally low-oncoprotein sample cannot distinguish 'IRX1 not occupied' from 'nothing occupied'.

Go deeper
A73051 beside the AFF1 HOXA-negative states
samplepartnerHOXAIRX1IRX1 callMEIS1 (2 kb)CDK6 (2 kb)depth
A73051AF100.87x0.25xNOT_OCCUPIED2.4/2.60.6/0.7ADEQUATE
p318AFF10.23x2.68xOCCUPIED22.1/12.816.3/9.8ADEQUATE
p454AFF10.33x6.35xOCCUPIED27.1/18.821.5/15.4ADEQUATE
TB920488AFF10.97x0.59xNOT_OCCUPIED5.9/11.25.4/23.2MARGINAL
TB986292AFF10.11x2.38xEQUIVOCAL10.0/7.712.7/40.4INADEQUATE

The comparison that carries the weight is A73051 against p318 and p454, all three depth-adequate and HOXA-negative: the two AFF1 blasts are IRX1-occupied and the AF10 blast is not. And TB920488 is the AFF1 counter-example in the same direction - HOXA-negative, IRX1-negative, with strong positive controls (MEIS1 5.9-11.2x, CDK6 5.4-23.2x), so HOXA-negative does not imply IRX1-positive even inside AFF1. Note that TB920488 is depth-MARGINAL (p_dz 0.135) and not adequate, correcting a premise carried into this beat.

What would make A73051 decisive rather than suggestive

A73051's own IRX1 transcript level. If IRX1 is expressed in A73051 and the oncoprotein is nonetheless absent from it, that separates the expression subgroup from the occupancy subgroup in a single sample and is the cleanest possible version of the result. GSE252378 has no RNA-seq, so this is not answerable inside the deposit.

Failing that: one more depth-adequate, positive-control-strong, HOXA-negative non-AFF1 sample. There is one in the deposit (A73051) and its controls are the weakest; the claim rests as much on the 17 HOXA-positive non-AFF1 samples, in which IRX1 is also absent, as on this one.

Facts and figures
sampleA73051
partnerKMT2A-AF10
lineagepaediatric AML
institutionFred Hutch Cancer Center
hoxa 20kb0.87
hoxa cluster0.701
hoxa9 promoter0.709
hoxa10 promoter0.885
irx1 gene0.249
irx1 promoter0
own floor0.255
p dz0.012
depthADEQUATE
readabilitypasses on MEIS1 only
meis1 promoter
[
  2.42,
  2.61
]
cdk6 promoter
[
  0.61,
  0.66
]
gapdh promoter
[
  3.53,
  2.45
]
analyses
[
  "ana-q147-irx1-fusion-occupancy-is-confined-to-kmt2a-aff1"
]
Still open
  1. Is IRX1 transcribed in A73051? That single RNA measurement would separate the expression subgroup from the occupancy subgroup definitively.
  2. ANSWERED in cycle 138, after two wrong turns. q148 said 'globally low-oncoprotein'; I called that unresolvable in cycle 135 on a depth confound; q150 showed the variable I called depth was the in-promoter ENRICHMENT fraction. On TRUE sequencing depth A73051 is at the 37th-39th percentile (9,101,995 and 9,279,052 fragments, ranks 20-21 of 54, median 10,300,517), and the share-versus-depth correlation is NEGATIVE (-0.255 over 54 libraries, -0.289 over 27 samples). So the low share is not a depth artefact and q148's answer stands: GLOBALLY LOW. The remaining caveat is enrichment, not depth, and the depth-matched down-sampling is no longer the discriminator it looked like.
Every source behind this entry (1)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    A73051 (KMT2A-AF10, paediatric AML, Fred Hutch) has KMT2A-N shares of 0.870x at the HOXA9-HOXA10 span, 0.701x over the 107 kb HOXA cluster, 0.249x at the IRX1 gene body against its own 0.255x floor, and exactly 0.000 at the IRX1 promoter in both N antibodies. Depth is ADEQUATE (p_dz = 0.0120). MEIS1 is 2.42x/2.61x and CDK6 is 0.61x/0.66x, so it passes the readability gate on MEIS1 alone.

    What this evidence cannot show

    One sample. Its positive controls are the weakest in the deposit, so a globally low-oncoprotein state cannot be excluded and would explain both its HOXA and its IRX1 readings without any locus selectivity. GAPDH at 3.5x/2.4x and PAX5 at 0.64x/0.96x are consistent with a working library at ordinary promoters, which argues against a pure sequencing failure, but not against a low-oncoprotein sample. 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 deposit's own paper reports that oncoprotein genomic enrichment varies widely between samples and that low-oncoprotein samples redirect to GMP genes rather than to no genes - so a low-HOXA sample may have a third programme neither HOXA nor IRX1, which this beat did not measure. Janssens DH et al. 2024
Connected to