The deposit paper's own supplementary table calls IRX1 an oncoprotein target in exactly p279, p318 and p454 - the same three samples this project reached by a different instrument
also called Janssens Supplementary Data 2 IRX1 row · q147 gene-scale IRX1 reproduced by the authors' own pipeline
This project decided which patients had the fusion protein sitting on the IRX1 gene using a simple rule of its own: twice the average gene switch, in two antibodies. The scientists who generated the data used a completely different method - peak calling with a statistical model - and published their answer in a supplementary spreadsheet. Both come out with the same three patients.
The same thing, for a scientist
Supplementary Data 2 of Janssens et al. 2024 (the GSE252378 paper) is a table of MLL-oncoprotein target genes called per sample: 1,091 named genes over 37 samples. Its IRX1 row reads exactly `['p279', 'p318', 'p454']` [E1] - identical to the sample set q147 derived at gene-body scale from a >=2.0x-of-an-average-promoter rule in two KMT2A-N antibodies plus a width-matched floor [E4]. The authors' pipeline shares none of that machinery: SEACR peak calls on pooled libraries with IgG peaks subtracted, then a Bayesian multinomial model of N-terminal versus C-terminal read distribution giving a z and a BH-corrected p per interval, with intervals passing N/C > 1.75 at p < 1e-5 called as oncoprotein target sites and pooled into a master list of 1,692 intervals [E3]. The main text states the conclusion in its own words: *"In a subset of the infant B-ALLs we profiled, MLL::AF4 binds the IRX1 locus at the expense of HOXA9, but MLL::ENL did not bind IRX1 in any of the samples we profiled"* [E2]. Two further cross-checks fall out of the same table: A73051 and 152985 - the two samples this project independently flagged, one as the HOXA-negative IRX1-negative AF10 blast and one as the drug-evicted library - carry **zero** target genes anywhere in the authors' table [E1].
What agreement between the two instruments does and does not buy
It buys a great deal about instrument artefact and nothing about sampling. Both readings are of the same 133 AutoCUT&RUN libraries from the same 27 patients, so a shared property of the material - a mapping artefact, a copy-number gain at 5p15, a batch effect - would move both. What it excludes is that the three-sample IRX1 call is a quirk of this project's 2.0x-plus-floor rule, which was the live worry going into this beat and is now closed 14.
It is also worth being exact about which claim is reproduced. The authors bound their negative to MLL::ENL - the partner they name - not to all five non-AFF1 partners. This project's 0 of 18 covers ENL, AFDN, AF10, AF9 and EPS15. The broader bound is this project's, not theirs 2.
The two instruments do not agree everywhere, and where they differ this project's rule is the permissive one
On the positive control they diverge. q147 calls HOXA9/HOXA10 occupied in 17 of the 18 non-AFF1 samples; the authors call HOXA9 a target in 11 of the same 18 14. On the reciprocal set, the authors' whole genome-wide call set is 1,091 genes, while this project's promoter rule passes a median 10.75% of 19,404 promoters per sample - roughly 2,086 calls per sample against the authors' ~125 1. The agreement at IRX1 is therefore agreement between a strict instrument and a loose one at a locus where the signal is large enough that the difference does not bite.
The published mechanism runs the other way round, and should not be read as support
The paper most often cited for IRX1 in t(4;11) is Kuhn, Loscher and Marschalek 2016, whose title calls the IRX1/HOXA connection a novel t(4;11)-specific cancer mechanism. Its abstract's mechanism is that IRX1 protein binds the MLL-AF4 complex at target gene promoters and counteracts its promoter-activating function 5 - IRX1 acting on the fusion, not the fusion sitting on IRX1. A title match is not a mechanism match, and this entry does not claim the 2016 paper supports fusion occupancy at the IRX1 locus.
| script | analyses/q154-gene-scale-occupancy-null/authors_null2.py |
|---|---|
| status | run |
| result summary | The authors' own published per-sample target table gives IRX1 = {p279, p318, p454}, identical to q147's gene-body call set derived by an unrelated rule. |
inputs
[ "Janssens 2024 Supplementary Data 2 (41467_2024_53399_MOESM4_ESM.xlsx) via Europe PMC supplementaryFiles for PMC11522475", "analyses/q147-hoxa-irx1-across-partners/results_gene_scale147.json" ]
authors irx1 samples
[ "p279", "p318", "p454" ]
authors irx2 samples
[ "p454" ]
genes called in exactly that sample set
[ "IRX1", "LCP1", "MAP3K1", "STARD3NL" ]
samples with zero target genes in the authors table
[ "A73051", "152985" ]
measurements — the numbers behind the chart
[
{
"quantity": "instrument_agreement_on_irx1_call_set",
"subject": "the set of GSE252378 samples in which IRX1 is called an oncoprotein target",
"value": 1,
"units": "Jaccard of the two call sets",
"n": 26,
"universe": "26 GSE252378 KMT2A-fusion patient samples (27 minus the drug-exposed 152985); this project's >=2.0x two-antibody gene-body rule versus the authors' SEACR + Bayesian N/C > 1.75 at p < 1e-5 pipeline",
"null": null,
"note": "both give exactly {p279, p318, p454}; same libraries, so this bounds instrument artefact, not sampling"
},
{
"quantity": "n_genes_with_the_identical_call_set",
"subject": "genes in the authors' 1,091-gene table called in exactly {p279, p318, p454}",
"value": 4,
"units": "genes",
"n": 1091,
"universe": "Janssens 2024 Supplementary Data 2, 1,091 oncoprotein-target genes across 37 samples",
"null": null,
"note": "IRX1, LCP1, MAP3K1, STARD3NL"
},
{
"quantity": "hoxa9_called_in_nonaff1_samples",
"subject": "HOXA9 called as an oncoprotein target by the authors' pipeline among the 18 non-AFF1 samples",
"value": 11,
"units": "samples",
"n": 18,
"universe": "the 18 non-AFF1 GSE252378 samples (19 minus the drug-exposed 152985)",
"null": null,
"note": "this project's 2.0x rule calls HOXA occupied in 17 of the same 18 - the project rule is the permissive one even on the positive control"
}
]- If a menin-inhibitor response biomarker were built on IRX1 occupancy, the agreement of two instruments on three samples still rests on three patients from one institution - two of whom are one patient sampled twice.
- The authors call IRX2 a target in p454 only. This project's promoter-scale q151 called IRX2 occupied in 0 of 27. Which is right, and does the disagreement come from the 2 kb promoter window missing a gene-body or enhancer signal that the peak caller finds?
- Only 4 of the 17 promoters in q152's AFF1-only recurrent core (BLNK, EPHA7, MATR3, WT1) appear anywhere in the authors' 1,091-gene table. Are the other 13 below the authors' threshold, or are they at intervals the authors' peak caller never nominated?
- The authors' master list is 1,692 intervals for the whole study. What is the corresponding number for this project's rule, and can the project's rule be re-thresholded to match it rather than being re-interpreted after the fact?
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Supplementary Data 2 of Janssens et al. 2024 is a table of 1,091 MLL-oncoprotein target genes with the list of samples each was called in, and its IRX1 row is exactly ['p279', 'p318', 'p454']; A73051 and 152985 appear in no row at all; HOXA9 is called in 19 of the 37 samples in the table and in 11 of the 18 non-AFF1 samples this project scores.
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 2024What this evidence cannot show
The row was read out of 41467_2024_53399_MOESM4_ESM.xlsx, fetched from the Europe PMC supplementaryFiles archive for PMC11522475 and parsed directly from the sheet XML. A full-text search of the article body will not find it, because the table is a spreadsheet. The sample naming in that table differs from GEO's for the St Jude series (TB-05-1939 there, TB051939 in the GEO titles); the mapping is by the shared digits and was checked for all seven TB samples.
Evidence that argues against this (1)
- This is the same deposit and the same libraries that q147 scored, so agreement between the two calls bounds instrument artefact only. It is not independent replication and nothing here adds a patient. NCBI Gene Expression Omnibus 2024
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The paper's discussion states that in a subset of the infant B-ALLs profiled, MLL::AF4 binds the IRX1 locus at the expense of HOXA9, while MLL::ENL did not bind IRX1 in any sample profiled - so the authors' own negative is bounded to ENL, not to all five non-AFF1 partners.
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 2024What this evidence cannot show
Read from the Europe PMC fullTextXML of PMC11522475 (81,046 characters after tag stripping). The word IRX1 occurs 10 times in the body; AFF1 occurs zero times - the paper writes MLL::AF4 throughout, so a query on AFF1 against this full text returns nothing.
Evidence that argues against this (1)
- The authors' statement is about MLL::ENL only. Reading it as support for a five-partner negative would overstate it; the five-partner bound is this project's own and rests on the same 18 libraries. NCBI Gene Expression Omnibus 2024
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The authors' target-calling pipeline is SEACR peak calling with IgG peaks subtracted, then a Bayesian model with two multinomial distributions over C-terminal and N-terminal reads giving a z and a Benjamini-Hochberg-corrected p per interval; intervals with N/C > 1.75 and p < 1e-5 were called as oncoprotein target sites, pooled into a master list of 1,692 intervals.
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 2024What this evidence cannot show
Methods section, read from full text. The 1,692 figure is the master list of genomic intervals; the gene-level table derived from it carries 1,091 named genes, so intervals and genes are not one-to-one.
Evidence that argues against this (1)
- An N/C ratio gate is itself a rule with a threshold, and the 1.75 and 1e-5 cut-points are the authors' choices. It is a different instrument from this project's, not a gold standard. Janssens DH et al. 2024
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
q147 independently called IRX1 gene-body OCCUPIED in exactly p279, p318 and p454 of the 8 KMT2A::AFF1 samples and in 0 of 18 non-AFF1 samples, with HOXA9/HOXA10 occupied in 17 of those 18.
Q147 gene-scale results - IRX1 and HOXA occupancy across six KMT2A fusion partners · Iqbal Lab oncofusion knowledge graph, beat Q147 · this project 2026What this evidence cannot show
q147 read bigWig mean coverage; this beat re-derived the same quantity from streamed fragment BEDs and reproduced the call set. Both are this project's own computation on the deposit.
Evidence that argues against this (1)
- q147's gene-scale call rests on a 2.0x gate whose floor clause is inert at that scale (0 of 27 libraries floor-limited), so it is a bare two-fold threshold with no per-library calibration. A different threshold would give a different sample set, and the agreement with the authors' pipeline is agreement at one particular cut-point. Iqbal Lab oncofusion knowledge graph et al. 2026
- E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The 2016 paper whose title calls the IRX1/HOXA connection a t(4;11)-specific mechanism reports IRX1 protein binding TO the MLL-AF4 complex at target gene promoters and counteracting its promoter-activating function - the opposite direction from fusion occupancy at the IRX1 locus.
The IRX1/HOXA connection: insights into a novel t(4;11)-specific cancer mechanism · Kuhn A, Loscher D, Marschalek R · Oncotarget 7:35341-52 2016What this evidence cannot show
Read from the efetch abstract. The paper is about IRX1 as an effector acting on the fusion complex and about HDAC inhibition; it contains no chromatin occupancy measurement at the IRX1 locus and no CUT&RUN.
Evidence that argues against this (1)
- The direction of the Kuhn mechanism does not make fusion occupancy at IRX1 less likely - a gene whose product feeds back on the complex may well also be a target of it - but it does mean the 2016 title cannot be quoted as prior evidence for the occupancy claim. Symeonidou V et al. 2021
- supports →IRX1 fusion occupancy is confined to KMT2A::AFF1: 0 of 18 samples across five other partners
the authors' own SEACR + Bayesian N/C pipeline returns the identical IRX1 sample set {p279, p318, p454}, so the call is not an artefact of this project's 2.0x rule
- related to →IRX1 — Iroquois homeobox 1, the gene that names the HOXA-negative half of infant t(4;11) ALL
IRX1 is an oncoprotein target gene in 3 of 37 samples in the deposit paper's own supplementary table
- related to →KMT2A::AFF1 (MLL-AF4), t(4;11)(q21;q23)
all three IRX1-positive samples carry KMT2A::AFF1; the authors bound their own negative to MLL::ENL only
- related to →IRX2 is not fusion-occupied at promoter scale in GSE252378: 0 of 27 samples, against a null of 2.73
the authors' table calls IRX2 a target in p454, where this project's promoter-scale rule calls it occupied in 0 of 27 - an unresolved instrument disagreement
the permutation is run over that table
same supplementary table
the table the cross-instrument audit reads