Oncofusion

start / graph / ana-q148-tb920488-the-other-double-negative-blast-has-the-highest-gmp-ratio-of-any-b-lineage-sample-and-does-not-survive-either-control

Analysismoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q148-what-the-hoxa-negative-blast-occupies

TB920488, the other HOXA-negative IRX1-negative blast, has the highest GMP-over-B-lineage ratio of any B-lineage sample in the deposit - and it fails both the matched null and the N-over-C test

The second of the two patients whose fusion protein is on neither usual gene set does look, on one measure, as though it has moved to myeloid genes. Two stricter checks say no.

The same thing, for a scientist

**TB920488** - KMT2A::AFF1, infant Pre-B-ALL, St. Jude, depth-MARGINAL - is the second instance of the double-negative state (HOXA **0.968x**, IRX1 **0.588x** in q147). Its `AML_specific` / `ALL_specific` ratio is **2.397**, which is **above every depth-ADEQUATE B-lineage sample in the deposit** (range 0.874-1.987) and above the B-lineage median of 1.268. Taken alone that is the GMP-DESTINATION phenotype in the one sample where it was not pre-specified to be looked for. **It does not survive either control.** - **Matched null:** `AML_specific` 1.884 against a coverability-matched null of **1.883/2.026**, ratio **0.929**, two-sided permutation p = 0.73 and 0.83 - not separated from an equally coverable promoter set. - **N-over-C:** `AML_specific` N/C = **1.112**, far below the deposit's own 1.75 fusion-specificity criterion; its canonical class is 1.494, also below. - And the ratio is denominator-driven again: its `ALL_specific` is **0.786**, the third lowest of the 23 untreated non-circular samples. Its canonical `Oncoprotein_Target` share is **1.720x**, **fourth lowest of 27**, so TB920488 points the same way as A73051: a low-oncoprotein blast, not a redirected one. Depth-MARGINAL means positive calls only are admissible from it - and the positive call (the high ratio) is the one the controls remove.

Go deeper
Why this sample is worth its own entry

It is the only sample in the deposit that produces the GMP-DESTINATION signature on the headline ratio without being the pair that defined the classes. If the beat had scored only the ratio - which is what the published framing invites - TB920488 would have been reported as a confirmation, in a B-lineage sample, in the index fusion, at an institution other than the one the reagent came from. That is a much more attractive result than the one that survives.

It is removed by two controls that are cheap and were both pre-specified in spirit: an absolute-floor-and-null requirement, and the deposit's own N-over-C criterion. Record the near-miss rather than only the verdict, because the next reader of this deposit will find TB920488 the same way.

TB920488 and A73051 side by side
TB920488A73051
partnerKMT2A::AFF1KMT2A-AF10
lineage / institutioninfant Pre-B-ALL, St. Judepaediatric AML, Fred Hutch
depth gradeMARGINAL (p_dz 0.135)ADEQUATE (p_dz 0.012)
HOXA (q147)0.968x0.870x
IRX1 (q147)0.588x0.249x
Oncoprotein_Target share1.720 (4th lowest of 27)1.208 (lowest of 25 untreated)
Oncoprotein_Target N/C1.4941.048
AML_specific share1.8841.752
AML_specific / matched null0.929 (ns)0.936 (ns)
AML_specific N/C1.1120.964
AML_specific / ALL_specific2.3973.493

Two patients, two partners, two lineages, two institutions, and the same shape: a low canonical programme, a GMP-like class indistinguishable from a matched null, and no N-terminal excess anywhere. n = 2 and no statistic is attached to that agreement.

Facts and figures
sampleTB920488
partnerKMT2A-AFF1
lineageinfant Pre-B-ALL
institutionSt. Jude Children's Hospital
depthMARGINAL
p dz0.135
hoxa q1470.968
irx1 q1470.588
aml over all2.397
aml specific share1.884
aml specific matched null ratio0.929
aml specific n over c1.112
oncoprotein target share1.72
oncoprotein target n over c1.494
all specific share0.786
scriptanalyses/q148-what-the-hoxa-negative-blast-occupies/analyze148.py
statusrun
result summarythe one non-circular sample with the GMP signature on the ratio; fails both controls
b lineage depth adequate range
[
  0.874,
  1.987
]
measurements
[
  {
    "quantity": "AML_over_ALL",
    "subject": "TB920488 (KMT2A::AFF1, HOXA-negative, St. Jude, depth-MARGINAL)",
    "value": 2.397,
    "units": "ratio",
    "n": 2,
    "universe": "40 AML_specific over 37 ALL_specific promoters, KMT2A-N1 and N2",
    "null": "depth-ADEQUATE B-lineage samples span 0.874-1.987, median 1.268",
    "note": "the highest B-lineage value in the deposit; removed by the matched null and by N/C"
  },
  {
    "quantity": "observed_over_matched_null",
    "subject": "TB920488 AML_specific class",
    "value": 0.929,
    "units": "ratio",
    "n": 2,
    "universe": "same promoters and libraries",
    "null": 1,
    "note": "two-sided permutation p 0.73 and 0.83; not separated from an equally coverable set"
  }
]
How this could bite us back
  • This is the result that would have been published if only the ratio had been scored: a confirmation in the index fusion, in a B-lineage sample, at a different institution from the reagent. Anyone reading the deposit will find it the same way.
Still open
  1. TB920488's single-locus positive controls (MEIS1 5.9-11.2x, CDK6 5.4-23.2x) and its 827-promoter class mean (1.720x) disagree about how much oncoprotein it carries. Which is right, and does a few-loci-strong, genome-wide-weak profile have a name?
  2. Would deeper sequencing of TB920488 move its ratio? It is the only sample where more depth could change a verdict in this beat, and it is the index fusion, which makes it the most valuable single library to re-sequence.
Every source behind this entry (2)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    TB920488's AML-over-B-lineage occupancy ratio of 2.397 exceeds every depth-adequate B-lineage sample in GSE252378 (range 0.874-1.987), but its GMP-like class is not separated from a coverability-matched null (0.929x, two-sided p 0.73 and 0.83) and its GMP-like N-over-C is 1.112.

    What this evidence cannot show

    TB920488 is depth-MARGINAL (p_dz 0.135), so only positive calls are admissible from it and a null result in it is weakly measured rather than absent. The comparison range excludes MARGINAL and INADEQUATE B-lineage samples, so TB920488 is being compared against a better-measured group than itself.

    Evidence that argues against this (1)
    • Comparing a depth-MARGINAL sample's ratio against a depth-ADEQUATE range is unfair in the direction that inflates it: a shallower library has more zeros in the denominator class, which raises the ratio mechanically. That alone could produce 2.397 without any redistribution. Q148 (Oncofusion Knowledge Graph) 2026
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    TB920488's canonical oncoprotein-target class sits at 1.720x, the fourth lowest of the 27 samples, so it is a low-oncoprotein sample like A73051.

    What this evidence cannot show

    Fourth lowest of 27 is not extreme, and the three below it are 152985 (drug-exposed), A73051 and A93949 (depth-MARGINAL). The untreated range is 1.208-3.330 so 1.720 is in the lower quarter, not an outlier.

    Evidence that argues against this (1)
    • The existing corpus entry for this sample records strong positive controls - MEIS1 5.9-11.2x and CDK6 5.4-23.2x - which is not the profile of a globally low-oncoprotein blast. Those are single loci and this is an 827-promoter class mean, and the two readings genuinely disagree about how much oncoprotein TB920488 has on chromatin. Q148 (Oncofusion Knowledge Graph) 2026
Connected to