Oncofusion

start / graph / absence-q149-the-occupancy-deposit-holds-no-megakaryoblastic-sample-so-the-zero-of-eighteen-cannot-cover-the-irx1-expressing-subgroup

Analysisstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q149-irx1-expression-across-partners

GSE252378 contains no megakaryoblastic sample, so the 0-of-18 non-AFF1 IRX1 occupancy bound does not cover the one subgroup where IRX1 is actually expressed

also called q149 the two populations barely intersect · occupancy cohort has no M7

The study that found the fusion protein never sits on IRX1 outside one partner did not include any child with the kind of leukaemia in which IRX1 is switched on. So it could not have seen the thing we now know to look for.

The same thing, for a scientist

**The occupancy result and the expression result are in populations that barely overlap, and this entry is the measurement of that.** `ana-q147-irx1-fusion-occupancy-is-confined-to-kmt2a-aff1` reports **0 of 18** non-AFF1 KMT2A-r samples with KMT2A-N oncoprotein at IRX1, HOXA occupied in 17 of the same 18, rule of three below 16.7%. It is a sound measurement. **Every disease label in GSE252378, read from the cached `!Sample_characteristics_ch1 = sample:` field, and there is no megakaryoblastic or AMKL sample among them** [E1]: | label as deposited | partners carrying it | |---|---| | `pediatric AML` | AFDN (4), AF10 (2), AF9 (1) | | `infant BALL` / `pediatric BALL` | ENL (3), AFF1 (3) | | `infant Pre-BALL` | AF10, AF9, AFF1 (4) | | `infant TALL` | ANL (1) | | `infant AML lineage-switching sample` | AFF1 (1) | | `infant B-ALL lineage-switching sample` | AFF1 (1) | A case-insensitive grep of the whole cached GSM record for `megakary|AMKL|erythro` returns **only** the four hits for the antibody clone name *CST D2M7U*, which contains the substring `M7`. **Verified-absent, not unchecked.** **Meanwhile `ana-q149-irx1-in-kmt2a-r-paediatric-aml-is-a-megakaryoblastic-lineage-property-not-a-fusion-property` places 13 of the 20 IRX1-expressing non-AFF1 KMT2A-r AML cases in FAB M7.** So the population in which IRX1 is on in KMT2A-r myeloid disease is precisely the population the occupancy deposit contains none of. **What this does and does not do to the q147 claim.** It does not refute it - the 0-of-18 is true of the 18 samples measured. It **bounds its scope**: the claim is *IRX1 is unoccupied in non-AFF1 KMT2A-r B-ALL, T-ALL and non-megakaryoblastic AML*, and the megakaryoblastic case has never been measured. **That subgroup is `never-measured`, not negative.**

Go deeper
The sharpest single instance: KMT2A::EPS15

The occupancy arm's EPS15 group is R3595 and R3652, with IRX1 within-sample shares of 0.109x and 0.047x against HOXA at 5.74x and 2.94x - deeply unoccupied. And q147's own manifest reading records that R3595 and R3652 are the same patient sampled twice, so the EPS15 occupancy arm is one patient.

In GSE68720, 2 of 3 KMT2A::EPS15 infant ALL cases carry IRX1 at 10.11 and 11.07 log2 against per-sample floors of 6.41 and 6.75 - as high as the highest KMT2A::AFF1 case in that cohort (11.37).

One patient with no oncoprotein at IRX1, and two different patients with IRX1 near the cohort maximum, in the same fusion. Different children, different assays, different cohorts - so this is not a paired demonstration of expression-without-occupancy. It is the reason exp-q149-paired-rna-and-oncoprotein-occupancy-in-a-single-kmt2a-eps15-blast is the cheapest decisive experiment in the whole line of work.

How the labels were read, so the absence can be re-checked

Source: analyses/is-there-a-menin-readout-under-dot1l-inhibition/gsm_GSE252378.txt, the cached GEO GSM brief-format record for GSE252378 - 7,120 lines, 171 GSMs.

Queries run, verbatim, with their results:

  • grep -c '^\^SAMPLE' -> 171
  • grep 'library_strategy' | sort | uniq -c -> 171 of 171 !Sample_library_strategy = OTHER
  • grep -ic 'RNA-seq|RNA-Seq|transcriptome|Expression profiling' -> 0. This independently confirms the premise that the deposit has no RNA.
  • grep -i 'megakary|AMKL|M7|FAB|erythro' -> the only hits are 4 antibody lines containing CST D2M7U, plus the per-sample sample: lines matched via the same clone string. No disease label contains a megakaryoblastic term.
  • The disease field is the third semicolon-delimited token of !Sample_characteristics_ch1 = sample: <antibody>; <sample id>; <fusion>; <disease>; <institution>, extracted for all 171 records.

The limitation of this bound: GEO's own free-text disease label is all there is. A sample labelled pediatric AML could be megakaryoblastic without GEO saying so - the deposit carries no FAB field. So the correct reading is the deposit does not state that any sample is megakaryoblastic, and a reader wanting certainty would have to ask the depositors. That is a weaker claim than none is, and it is the one made here.

Facts and figures
absence classnever-measured
what is absentoncoprotein occupancy at IRX1 in any megakaryoblastic KMT2A-rearranged blast
deposit checkedGSE252378
n gsm171
rna mentions in record0
megakaryoblastic labels present0
library strategy values
{
  "OTHER": 171
}
disease labels present
[
  "pediatric AML",
  "infant BALL",
  "pediatric BALL",
  "infant Pre-BALL",
  "infant TALL",
  "infant AML lineage-switching sample",
  "infant B-ALL lineage-switching sample"
]
eps15 occupancy arm
{
  "samples": [
    "R3595",
    "R3652"
  ],
  "same_patient": true,
  "irx1_share": {
    "R3595": 0.1091,
    "R3652": 0.0472
  },
  "hoxa_share": {
    "R3595": 5.7389,
    "R3652": 2.9423
  }
}
eps15 expression arm GSE68720
{
  "n": 3,
  "irx1_values": [
    3.24,
    10.11,
    11.07
  ],
  "floors": [
    null,
    6.41,
    6.75
  ]
}
measurements
[
  {
    "quantity": "count_megakaryoblastic_samples",
    "subject": "GSE252378 disease labels, all 171 GSMs",
    "value": 0,
    "units": "samples",
    "n": 171,
    "universe": "case-insensitive grep for megakary|AMKL|erythro over the cached GSM brief record; only hits are the antibody clone string CST D2M7U",
    "null": null,
    "note": "the deposit carries no FAB field, so this bounds what the deposit STATES, not the true phenotypes"
  },
  {
    "quantity": "n_independent_patients_in_EPS15_occupancy_arm",
    "subject": "GSE252378 KMT2A-EPS15",
    "value": 1,
    "units": "patients",
    "n": 2,
    "universe": "R3595 and R3652; q147's manifest reading records them as the same patient sampled twice",
    "null": null,
    "note": "the partner with the highest IRX1 expression rate in infant ALL has an occupancy arm of one patient"
  }
]
How this could bite us back
  • Any decision to deprioritise IRX1 as a target outside KMT2A::AFF1 would rest on a bound that structurally cannot see the megakaryoblastic subgroup. Deciding not to look is the expensive error here, because the subgroup is 20 of 171 KMT2A-r AML cases and 65% of them are IRX1-positive.
Still open
  1. Does any deposited oncoprotein occupancy dataset anywhere contain a megakaryoblastic KMT2A-rearranged blast? If one does, the single most informative IRX1 measurement in this line of work is already public and unanalysed.
  2. Are any of the seven GSE252378 samples labelled only 'pediatric AML' in fact megakaryoblastic? The depositors know; the deposit does not say, and one email would resolve the strongest scope limit on the q147 result.
Every source behind this entry (2)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

    The 171 GSM records of GSE252378 carry disease labels of pediatric AML, infant and pediatric B-ALL, infant Pre-B-ALL, infant T-ALL and two lineage-switching samples, and no label contains a megakaryoblastic, AMKL or erythroid term; all 171 records also carry library_strategy OTHER and the record contains no mention of RNA-seq or transcriptome.

    What this evidence cannot show

    Bounded by the depositors' free-text disease labels, which are all the deposit provides - there is no FAB field, no immunophenotype and no blast count. A sample labelled 'pediatric AML' could be megakaryoblastic without the label saying so, so this establishes that the deposit does not STATE any megakaryoblastic sample, not that none is. Read from a locally cached copy of the GSM records rather than re-fetched this session. The grep for M7 also matches the antibody clone name CST D2M7U, which is why the query was widened to megakary/AMKL/erythro.

    Evidence that argues against this (1)
    • Two GSE252378 samples are explicitly LINEAGE-SWITCHING (p279 B-ALL-to-AML, p318 infant AML), and lineage switch under KMT2A-r disease can go to a monocytic or occasionally a megakaryocytic phenotype. Both are KMT2A::AFF1 and both are in the IRX1-OCCUPIED set, so they do not populate the non-AFF1 arm - but they are the samples in the deposit whose lineage is least well described by a single label, and a reader should not treat the label list as a clean lineage census. GEO depositors et al. 2024
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    Thirteen of the twenty IRX1-expressing non-AFF1 KMT2A-rearranged paediatric AML cases are FAB M7, so the population in which IRX1 is transcribed in KMT2A-rearranged myeloid disease is the one the occupancy deposit does not contain.

    A new genomic framework to categorize pediatric acute myeloid leukemia · Umeda M, Ma J, Westover T, et al. · Nat Genet 56:281-293 2024
    PMID 38212634DOI
    What this evidence cannot show

    The two cohorts are different institutions, different patients and different assays; no child appears in both. The M7 concentration is itself an association with morphology and not with the fusion - 54 of 97 non-KMT2A-r M7 cases are also IRX1-positive - so the subgroup the occupancy deposit lacks is a morphological one, and the scope limit this entry places on q147 is about lineage coverage rather than about partner coverage.

    Evidence that argues against this (1)
    • If the IRX1 signal in M7 comes from normal megakaryocyte-erythroid cells in the sample rather than from blasts, then the subgroup the occupancy deposit lacks would not be a leukaemic IRX1 subgroup at all, and the scope limit placed on q147 here would be unnecessary. No purity measurement exists for either cohort. Nagel S et al. 2022
Connected to