Oncofusion

start / graph / finding-q154-irx2-is-fusion-occupied-at-gene-body-scale-in-p454-and-the-promoter-scale-zero-was-a-knife-edge

Analysismoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q154-gene-scale-occupancy-null

IRX2 IS fusion-occupied at gene-body scale in p454 - and q151's promoter-scale zero of 27 turns on one replicate 5% below the threshold, which two other instruments put above it

also called q154 IRX2 gene-body measurement · closing the q153 IRX2 gene-body absence

An earlier round found no sign of the fusion protein on IRX2 in any of the twenty-seven patients, using a two-kilobase window at the gene's start. Measuring the whole gene instead, one patient - p454 - is clearly positive, and the original scientists' own published list also names IRX2 in that one patient. The earlier zero came down to a single measurement of 1.898 against a cut-off of 2.0.

The same thing, for a scientist

**This closes `absence-no-gene-body-scale-irx2-measurement-exists-in-gse252378`.** A genome-wide gene-body matrix over all 133 GSE252378 libraries puts IRX2 (chr5:2,745,844-2,751,677, 5,833 bp, hg38) at **5.145x and 3.575x an average promoter in the two KMT2A-N libraries of p454**, clearing the 2.0x gate and its floor in both, and **occupied in 1 of 27 samples - 1 of 8 AFF1, 0 of 18 non-AFF1** [E1]. **Janssens et al.'s own Supplementary Data 2 independently lists IRX2's oncoprotein-target sample set as exactly `['p454']`** [E2]. The promoter-scale negative that q151 recorded is more fragile than it reads. Re-derived from the same bigWig matrix q151 used, p454's IRX2 promoter shares are **3.812x and 1.898x** against a floor of 0.015 and 0.202 - so the call fails on **one replicate, by 0.102, a 5.1% margin** [E3]. Re-measured from streamed fragment midpoints over the identical 2 kb interval, the same two libraries give **4.589x and 2.351x** and the promoter **passes** [E1]. IRX1's promoter calls agree exactly between the two instruments (2 of 27, p279 and p454), so this is not a general disagreement - it is a threshold crossing at one locus in one replicate [E1][E3]. **q151's claim should be read as "IRX2 does not clear a 2.0x promoter gate in the bigWig matrix", not as "IRX2 is not fusion-occupied".**

Go deeper
The three instruments at IRX2 in p454
instrumentintervalreplicate 1replicate 2floorcall
bigWig mean coverage (q151/q152's matrix)2 kb promoter3.8121.8980.015 / 0.202fails by 0.102 in rep 2
streamed fragment midpoints (this beat)the same 2 kb promoter4.5892.3510.046 / 0.251passes
streamed fragment midpoints (this beat)5,833 bp gene body5.1453.575per-library max of 8 negativespasses
Janssens et al. SEACR + Bayesian N/C > 1.75, p < 1e-5called peak intervals--IgG-subtractedcalled a target in p454

No other sample of the 27 is close at either scale, so the 1-of-27 count is not itself fragile - only p454's inclusion is, and three of the four readings include it 123.

Why a 5% margin matters more here than usual

q151's IRX2 zero has been used as a specificity control for the IRX1 claim: IRX1 is occupied and its neighbour on the same chromosome arm is not, which argues the IRX1 signal is not an IRX-family or a 5p15 regional effect. That argument weakens if IRX2 is occupied in one AFF1 sample and no other. It does not disappear - IRX1 is 3 of 27 and IRX2 is 1 of 27, and IRX1's magnitudes are larger - but the clean contrast becomes a quantitative one.

It also means q153's downstream reasoning, which established that IRX2's promoter window is shared with IRX2-DT and asked whether an IRX2 measurement could be attributed at all, was asking the right question about a locus that does carry signal. The gene-body interval used here is IRX2's own transcript body and contains no other scored gene body; IRX2-DT has no row in this project's 19,404-symbol table at all 1.

Facts and figures
scriptanalyses/q154-gene-scale-occupancy-null/irx2_check.py
statusrun
result summaryIRX2 gene body occupied in p454 only (5.145x, 3.575x); authors' own table agrees; q151's promoter-scale zero rests on a single replicate at 1.898 against a 2.0 gate.
irx2 interval hg38chr5:2745844-2751677
irx2 length bp5833
closes absence entryabsence-no-gene-body-scale-irx2-measurement-exists-in-gse252378
inputs
[
  "GSE252378 GSM*_hg38.bed.gz (133 libraries, streamed)",
  "C:/Temp/q145irx1/promoters.tsv and refGene.txt.gz (hg38)",
  "C:/Temp/q147part/promoter_matrix.tsv (RAW mean bigWig coverage, 19,404 x 133)",
  "Janssens 2024 Supplementary Data 2"
]
irx2 occupied samples gene body
[
  "p454"
]
irx2 occupied samples authors table
[
  "p454"
]
measurements
[
  {
    "quantity": "within_sample_share_gene_body",
    "subject": "IRX2 gene body in p454, KMT2A-N replicates 1 and 2",
    "value": 5.145,
    "units": "multiples of an average promoter in the same library",
    "n": 2,
    "universe": "19,402 hg38 refGene gene bodies x 27 GSE252378 samples; replicate 2 is 3.575",
    "null": 0.608,
    "note": "the null quoted is the per-sample gene-body pass rate as a percentage, not a share; IRX2 is 1 of 27 samples occupied"
  },
  {
    "quantity": "margin_by_which_the_promoter_scale_negative_holds",
    "subject": "IRX2 promoter, p454, KMT2A-N replicate 2, in the bigWig matrix q151 used",
    "value": 1.898,
    "units": "multiples of an average promoter; the gate is 2.0",
    "n": 1,
    "universe": "19,404 2 kb promoter windows x 133 GSE252378 libraries, raw bigWig mean coverage converted to within-sample shares",
    "null": 2,
    "note": "fails by 0.102, a 5.1% margin; the fragment-midpoint instrument gives 2.351 for the same window and passes"
  }
]
How this could bite us back
  • If IRX2 as well as IRX1 is a KMT2A::AFF1 address, a therapeutic argument built on IRX1 alone targets one member of a family whose other member is also engaged - and IRX family members are known to compensate for one another in development.
Still open
  1. Is p454's IRX2 signal a separate address or the edge of the IRX1 domain? The two TSSs are 850 kb apart on 5p15, which is far for a single domain, but no intermediate interval has been scored.
  2. IRX2-DT has no row in this project's 19,404-symbol promoter table. q153 established that IRX2's promoter window is shared with it; at gene-body scale the two transcripts are separable, so a directed re-measurement could settle the attribution q153 left open.
  3. Which other promoter-scale zeros in q151 and q152 fail by less than 10%? A margin column on every negative call would say at once which of this project's absences are knife-edge and which are robust.
Every source behind this entry (3)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

    In a genome-wide gene-body matrix built from streamed GSE252378 fragment BEDs, IRX2 (chr5:2,745,844-2,751,677, 5,833 bp) is occupied by the two-antibody 2.0x-plus-floor rule in 1 of 27 samples - p454, at 5.145x and 3.575x - and in 0 of 18 non-AFF1; the same fragment instrument at the 2 kb promoter gives 4.589x and 2.351x in p454, and reproduces q151/q152's IRX1 promoter call of 2 of 27 exactly.

    What this evidence cannot show

    Fragment-midpoint counting is not bigWig mean coverage: a fragment that overlaps an interval's edge contributes fully to one and partially to the other. Over 21 named regions the two agree at Spearman 0.9932 with a per-region median ratio of 1.038, but at IRX2's promoter in one replicate they differ by 24%, which is enough to cross a 2.0 threshold. Neither is the ground truth.

    Evidence that argues against this (1)
    • One sample is one sample. p454 is a single Texas Children's infant B-ALL, and IRX2 at 1 of 27 is exactly the frequency at which a single-library artefact - a copy-number gain at 5p15, a mapping pile-up, a PCR duplicate cluster - would present. Nothing here excludes that, and no copy-number data exists for these samples in this deposit. NCBI Gene Expression Omnibus 2024
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Janssens et al.'s Supplementary Data 2 lists IRX2's oncoprotein-target sample set as exactly ['p454'], independently of this project's rule.

    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 · Nature Communications 15:9341 2024
    PMID 39472576DOI
    What this evidence cannot show

    The paper's body never discusses IRX2; the call exists only as a row in the supplementary spreadsheet, so the authors themselves attach no interpretation to it.

    Evidence that argues against this (1)
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

    Re-derived from C:/Temp/q147part/promoter_matrix.tsv - the raw bigWig mean-coverage matrix q151 and q152 used, converted to within-sample shares by dividing by each column's mean - p454's IRX2 promoter shares are 3.812 and 1.898 against floors of 0.015 and 0.202, so the promoter-scale negative fails on replicate 2 by 0.102, a 5.1% margin.

    Q154 re-derivation of the q151 IRX2 promoter call from the bigWig matrix · Iqbal Lab oncofusion knowledge graph, beat Q154 · this project 2026
    What this evidence cannot show

    The matrix holds RAW mean coverage, not shares; column means span 0.9177-18.7951 and 0 of 133 columns is within 1% of 1.0. The shares above were produced by dividing by the column mean, which is the conversion q152 used and asserted. A reader taking the raw values as shares would get a different answer at every absolute threshold.

    Evidence that argues against this (1)
    • A single replicate 5% below a threshold is not evidence that the threshold is wrong. q151's rule was pre-specified and applied consistently; reporting the margin is a statement about fragility, not about error, and a beat that re-examines only the calls it dislikes will find fragility everywhere. Iqbal Lab oncofusion knowledge graph et al. 2026
Connected to