KMT2A-N/C CUT&RUN on one untreated KMT2A::AFF1 blast from outside Texas Children's, scored at the six AFF1-exclusive gene bodies with a pre-declared margin column
also called the eighth AFF1 patient · q154 decisive experiment
Every patient in whom the fusion protein is found on IRX1 comes from one hospital, and two of the three are the same child sampled twice. One more untreated AFF1 patient from a different hospital, profiled the same way, would settle whether this is biology or a local effect.
The same thing, for a scientist
**The whole IRX1 gene-body result rests on `p279`, `p318` and `p454` - three samples, two patients, one institution (Texas Children's).** Holding institution fixed, IRX1's own label-permutation p is 0.119 (exact, over all 4,410 assignments); at patient level it is 0.077. **Neither clears 0.05 and one added independent AFF1 patient is enough to move both.** The experiment is one AutoCUT&RUN run: KMT2A-N (two antibodies), KMT2A-C (two antibodies) and IgG on a single untreated diagnostic KMT2A::AFF1 infant B-ALL blast from any centre other than Texas Children's, processed to hg38 fragment BEDs.
Design
- Material. One untreated diagnostic KMT2A::AFF1 infant B-ALL bone marrow or peripheral blood blast preparation, cryopreserved is acceptable, from a centre other than Texas Children's. No menin inhibitor exposure; the deposit already shows what that does.
- Assay. AutoCUT&RUN exactly as in GSE252378: KMT2A-N1, KMT2A-N2, KMT2A-C1, KMT2A-C2 and pooled IgG, paired-end, aligned to hg38, deposited as fragment BEDs so the existing extractor reads them unchanged.
- Depth. >= 8 M mapped fragments per library. The deposit's median is 9.9 M and its two shallowest KMT2A-N libraries (TB920488 at 0.46% and TB000694 at 0.18% call rate) are the ones that lose calls.
- Scoring. The published pipeline, unchanged: within-sample share = interval fragment density over that library's own mean promoter density, gate 2.0x in both KMT2A-N libraries, per-library floor from ALB, GFAP, INS, KRT14, MYH7, MYOD1, NEUROD1, RHO.
Decision rule, written before the sample is run
The six gene bodies are ELF1, GNAQ, H3C15, IRX1, PAN3, SPAG7, and the primary readout is IRX1.
| outcome at IRX1 | what it means |
|---|---|
| occupied (>=2.0x in both N libraries, above floor) | the address is not an institution effect. IRX1 becomes 4 of 9 samples and 3 of 8 patients; the patient-level permutation p falls from 0.077 to about 0.02 and the institution-stratified test is no longer degenerate. |
| at background (<1.0x in both, as all 18 non-AFF1 samples are) | AFF1-exclusivity survives but AFF1-recurrence weakens: 3 of 9 samples, 2 of 8 patients, and the claim becomes about a subgroup of AFF1 rather than about AFF1. |
| equivocal (1.0-2.0x) | the gradient model gains: IRX1 joins the five class members whose non-index samples sit under the bar, and the one clean margin in the class is lost. |
A margin column is mandatory in the report: the value at each of the six, not only the call. That single column is what separated IRX1 from the other five here and it costs nothing.
Why this rather than the alternatives
Cheaper than a partner swap. exp-q147-swap-the-aff1-c-terminus-onto-kmt2a-af9-and-score-irx1-occupancy tests mechanism and is the better experiment scientifically, but it needs a construct, a transduction model and a permissive cell of origin. This needs one patient sample and one CUT&RUN plate.
Nothing in the public archive substitutes for it. A GEO survey for a second deposit carrying per-patient KMT2A oncoprotein occupancy in KMT2A::AFF1 returned nothing usable: 17 series match KMT2A with CUT&RUN, and the ones that are on topic are NUP98-rearranged leukaemia, NPM1c, SETBP1-mutant leukaemia, T-ALL LEDGF and a Wiedemann-Steiner mouse. GSE252378 remains the only deposit of its kind, so the institution confound cannot be broken by re-analysis.
It also answers the five. If ELF1, GNAQ, PAN3, H3C15 and SPAG7 are gradient artefacts, a second AFF1 sample of typical rather than exceptional signal should fail to call them while still calling IRX1.
| design | One untreated diagnostic KMT2A::AFF1 infant B-ALL blast from a centre other than Texas Children's; AutoCUT&RUN with KMT2A-N1, KMT2A-N2, KMT2A-C1, KMT2A-C2 and pooled IgG; paired-end, hg38, deposited as fragment BEDs. |
|---|---|
| readout | Within-sample share at ELF1, GNAQ, H3C15, IRX1, PAN3 and SPAG7 gene bodies, reported as a value and not only as a call, plus the per-library floor from the eight cannot-be-occupied loci and the genome-wide pass rate. |
| controls | HOXA9/HOXA10 20 kb as the positive control (occupied in 17 of 18 non-AFF1 samples in the existing deposit); the eight-locus cannot-be-occupied panel as the floor; IgG recorded but not used as the gate, since IgG libraries in this deposit are 57-94% zero at promoters. |
| critical requirement | >= 8 M mapped fragments per KMT2A-N library, and a genome-wide gene-body pass rate within the deposit's observed 0.175-0.974% range - a sample outside that range is not comparable and its calls must not be pooled. |
| decision rule | IRX1 occupied -> the address is not an institution effect and the patient-level permutation p falls from 0.077 to about 0.02. IRX1 below 1.0x -> exclusivity survives, recurrence weakens to 2 of 8 patients. IRX1 between 1.0x and 2.0x -> the gradient model gains and the class loses its one clean margin. |
| evidence asymmetry | A positive result is decisive for the address; a negative is not decisive against it, because a single AFF1 patient can be HOXA-high and IRX1-low - three of the eight existing AFF1 samples are. |
| cost tier | low |
| timeline | weeks, given access to one banked sample |
tests hypotheses
[ "hyp-q154-gene-body-partner-exclusivity-is-a-threshold-on-an-oncoprotein-level-gradient-and-irx1-is-the-exception", "hyp-q147-the-irx1-address-requires-the-aff1-partner-not-the-menin-interface" ]
analyses
[ "finding-q154-irx1s-own-three-sample-pattern-does-not-clear-0-05-once-institution-or-lineage-is-held-fixed", "finding-q154-irx1-is-the-only-one-of-the-six-aff1-exclusive-gene-bodies-with-a-clean-margin-and-an-independent-confirmation" ]
measurements
[
{
"quantity": "n_independent_patients_currently_supporting_the_irx1_address",
"subject": "KMT2A::AFF1 patients in which IRX1 is fusion-occupied at gene-body scale",
"value": 2,
"units": "patients",
"n": 7,
"universe": "the 7 independent KMT2A::AFF1 patients in GSE252378 (8 samples, p279 and p318 being one patient), all IRX1-positive samples from Texas Children's",
"null": null,
"note": "patient-level label-permutation p 0.077; exact institution-stratified p 0.119"
}
]- If one added patient is IRX1-negative, the honest conclusion is that IRX1 occupancy marks a subgroup within KMT2A::AFF1 rather than the fusion - which is what the expression literature has said since 2016 and would make the occupancy claim a smaller result than it currently reads.
- Do the Janssens, Meshinchi or Mullighan groups hold further untreated KMT2A::AFF1 samples already profiled but not deposited? One email is cheaper than one CUT&RUN.
- Should the same sample also be profiled for IRX1 RNA? GSE252378 has no RNA-seq, so no sample anywhere pairs IRX1 occupancy with IRX1 transcript in the same cells, and that pairing is what decides whether the occupancy is functional.
- Is there a banked KMT2A::AFF1 sample from a HOXA-high, IRX1-low patient that could be run alongside? Two samples chosen to span the HOXA/IRX1 axis would test the reciprocal claim rather than only adding an n.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
All three IRX1-positive samples are Texas Children's, two of them are one patient, and holding institution fixed gives an exact label-permutation p of 0.119 for IRX1's pattern while patient collapse gives 0.077 - so one additional independent AFF1 patient is the smallest addition that changes the inference.
Q154 label permutation over the published GSE252378 oncoprotein-target call table · Iqbal Lab oncofusion knowledge graph, beat Q154 · this project 2026What this evidence cannot show
The projected p of about 0.02 for a positive ninth sample assumes the added patient is from a fourth institution and that no non-AFF1 sample is added with it. Adding a positive AFF1 sample AND a negative non-AFF1 sample from the same centre would move it further.
Evidence that argues against this (1)
- One added patient cannot separate institution from partner if that patient's material is handled by the same laboratory that generated the Texas samples. The Fred Hutch group processed all of GSE252378, so 'a different institution' must mean a different tissue source, not necessarily a different assay laboratory - and the confound this experiment addresses is the source, not the bench. NCBI Gene Expression Omnibus 2024
- E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database
A GEO series search for KMT2A or MLL-AF4 with CUT&RUN or ChIP-seq returns 17 and 2 series respectively, and none carries per-patient KMT2A oncoprotein occupancy in KMT2A::AFF1 outside GSE252378; the on-topic hits are NUP98-rearranged leukaemia, NPM1c, SETBP1-mutant leukaemia, a T-ALL LEDGF study and a Wiedemann-Steiner mouse model.
NCBI GEO DataSets search, 2026-09-24 · NCBI Gene Expression Omnibus · NCBI GEO 2026What this evidence cannot show
The esearch querytranslation shows `CUT&RUN` parsed as `CUT AND RUN`, so the query is looser than written and the 17 is an over-count rather than an under-count. The search was over series titles and summaries only; a series whose summary does not name the assay or the gene would be missed, and no data-availability statements were read. This is a bounded absence over GEO series metadata, not over the literature.
Evidence that argues against this (1)
- A metadata search cannot see a deposit whose summary describes the disease and not the assay, and this project has twice found that an absence of linkage in an archive is not an absence of data. Before commissioning wet work, the Janssens group and the St Jude and Texas Children's sample sources should be asked directly whether further AFF1 samples were profiled and not deposited. Janssens DH et al. 2024
one added AFF1 patient scores all six class members with a mandatory margin column
breaks the institution confound that currently holds IRX1's own p at 0.119
cheaper and less informative: one patient sample against a construct, a transduction model and a permissive cell of origin