HOXA occupancy is the partner-invariant: occupied in 17 of 18 non-AFF1 KMT2A-r samples, 2.94x to 29.1x
Whichever gene KMT2A is fused to, the fusion protein sits on the HOXA9/HOXA10 genes. That part of the disease is shared across all six partners tested.
The same thing, for a scientist
🟠 **CORRECTED 2026-09-24 (cycle 135): the OCCUPIED range is 2.94x-29.1x, not 0.87x-29.1x.** **0.87x is `A73051` - the single non-AFF1 sample that is NOT occupied.** Quoting it as the floor of the occupied range folded the one negative into the positives and made the pre-registered **>=2.0x** rule look self-contradictory, since no sample can be "occupied" at 0.87x. Re-derived from this beat's own `results_gene_scale147.json`: of the **18** untreated non-AFF1 samples, **17 are occupied spanning 2.942x (`R3652`) to 29.108x (`TB040576`)**, and `A73051` sits at **0.870x**, unoccupied. **The all-cohort range over all 18 non-AFF1 samples is 0.87x-29.11x and that figure is fine** - it is only wrong when labelled as the occupied range. Found by q148; the coordinator had repeated the error into the cycle-131 log, the ledger and two dispatch briefs. --- Across GSE252378's six fusion partners, **HOXA9/HOXA10 is fusion-OCCUPIED in 17 of 18 untreated non-AFF1 samples** at gene-body scale, spanning **2.94x to 29.1x** an average promoter [E1]. Occupied in KMT2A-ENL 6/6, -AFDN 4/4, -AF9 3/3, -EPS15 2/2, -AF10 2/3. By contrast HOXA is occupied in only **4 of 8** KMT2A::AFF1 samples, which is the lower rate in the deposit [E1] - the HOXA-dead states q145 found are, in this larger comparison, an **AFF1 peculiarity rather than a general KMT2A-r phenotype**. This is the substrate on which the menin-interface hypothesis sits most comfortably: the KMT2A N-terminal fragment retained in every one of these six fusions carries the menin and LEDGF interfaces and the CXXC domain, and it is the HOXA address - not the IRX1 address - that travels with it [E2].
The table, per partner
Untreated samples only; the one revumenib-exposed sample (152985) is excluded and reported separately. Values are gene-body-scale within-sample shares over the 17,883 bp HOXA9-HOXA10 span, mean of KMT2A-N1 and N2.
| partner | n | HOXA occupied | HOXA share range |
|---|---|---|---|
| KMT2A-AF9 | 3 | 3/3 | 5.69 - 29.11x |
| KMT2A-AFDN | 4 | 4/4 | 11.03 - 21.63x |
| KMT2A-ENL | 6 | 6/6 | 7.10 - 17.65x |
| KMT2A-EPS15 | 2 | 2/2 | 2.94 - 5.74x |
| KMT2A-AF10 | 3 | 2/3 | 0.87 - 20.17x |
| KMT2A-AFF1 | 8 | 4/8 | 0.03 - 11.78x |
The two lowest partner-level maxima are EPS15 (5.74x, n = 2) and AFF1. EPS15 is the one partner in the panel that is not an elongation-complex component, and its two samples are the lowest non-AFF1 HOXA values in the deposit - suggestive, but n = 2 and this cannot support a claim.
Why the AFF1 group is the odd one out and what that costs the original reading
q145 read three HOXA-dead KMT2A::AFF1 states as one arm of a two-arm split. That reading survives within AFF1. What this larger comparison adds is that HOXA-dead is rare outside AFF1: 1 of 18 non-AFF1 samples (A73051, which is also IRX1-negative), against 4 of 8 within AFF1.
Two readings, and this deposit cannot separate them:
- AFF1 fusions genuinely tolerate a HOXA-off state that the other partners do not.
- The AFF1 arm of this deposit is enriched for lineage-switch and low-oncoprotein samples - the deposit paper's own finding is that lineage-switching samples have reduced oncoprotein levels and redirect to GMP genes - and the HOXA-dead states are that, not a partner effect.
Reading 2 is the more parsimonious and is the one the deposit paper itself would predict.
| hoxa interval | chr7:27,159,437-27,177,320 (hg38, 17,883 bp, HOXA9-HOXA10 span) |
|---|---|
| hoxa occupied nonaff1 | 17/18 |
| hoxa occupied aff1 | 4/8 |
| corrected cycle 135 | range_nonaff1 was [0.87, 29.108] and was quoted as the OCCUPIED range; 0.87 is A73051, the one unoccupied sample. Split into range_occupied_nonaff1 [2.942, 29.108] and range_all_nonaff1 [0.87, 29.108] so the ambiguous key cannot be requoted. |
per partner occupied
{
"KMT2A-ENL": "6/6",
"KMT2A-AFDN": "4/4",
"KMT2A-AF9": "3/3",
"KMT2A-EPS15": "2/2",
"KMT2A-AF10": "2/3",
"KMT2A-AFF1": "4/8"
}analyses
[ "ana-q147-irx1-fusion-occupancy-is-confined-to-kmt2a-aff1" ]
range occupied nonaff1
[ 2.942, 29.108 ]
range all nonaff1
[ 0.87, 29.108 ]
- If HOXA is the partner-invariant address and it is menin-dependent, then a menin inhibitor should act on it in all six partners - and the escape route is whatever is NOT on that address, which in this deposit means the AFF1-restricted IRX1 programme and the lineage-switch GMP programme the deposit paper describes.
- HOXA9/HOXA10 are required for normal haematopoietic stem and progenitor function, so any strategy that works by shutting the shared HOXA address carries an on-target marrow cost that does not depend on which partner the patient has.
- Is the AFF1 group's low HOXA occupancy rate a partner property or a sampling property? The test is an AFF1 cohort not selected for lineage switching, matched to the non-AFF1 samples on oncoprotein level.
- KMT2A-EPS15 gives the two lowest non-AFF1 HOXA values and EPS15 is the one non-elongation partner in the panel. At n = 2 this is unquotable; does it hold in more EPS15 cases?
- E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
In GSE252378, KMT2A-N occupancy at the HOXA9-HOXA10 span clears the pre-registered absolute floor in 17 of 18 untreated non-AFF1 samples (ENL 6/6, AFDN 4/4, AF9 3/3, EPS15 2/2, AF10 2/3) at 2.94-29.1x an average promoter, and in only 4 of 8 KMT2A::AFF1 samples.
q147 re-analysis of GSE252378: HOXA and IRX1 oncoprotein occupancy across all six KMT2A fusion partners, 27 samples, 133 libraries (analyses/q147-hoxa-irx1-across-partners) · Iqbal Lab oncofusion knowledge graph, beat q147 · internal analysis over deposited GSE252378 bigWigs 2026What this evidence cannot show
One deposit, one assay. The AFF1 4/8 figure is not independent of q145, which selected this locus pair for study precisely because three AFF1 samples were HOXA-dead - so the AFF1 denominator is the one group whose composition was known before the comparison was designed, and the contrast between 4/8 and 17/18 is therefore partly a selection effect and must not be quoted as an unbiased partner comparison. Every number is a WITHIN-SAMPLE share: an interval's mean coverage divided by that same library's mean coverage over all 19,404 hg38 RefSeq promoters, so 1.0 = an average promoter in that library. GSE252378 has no spike-in, so no between-library absolute statement is licensed; the only cross-sample quantity used is a ratio of two loci measured inside one library, where the unknown per-library scale cancels. The occupancy call is absolute, not rank-based: >=2.0x an average promoter in BOTH KMT2A-N antibodies AND above that library's own floor, where the floor is the 95th percentile of eight width-matched windows at loci that cannot be occupied in a leukaemia (ALB, RHO, MYH7, MYOD1, NEUROD1, KRT14, GFAP, INS). IgG is not the gate: q145 measured GSE252378 IgG libraries as 57-94% zero at promoters, and one of them (TB051939) reads 13.1x an average promoter at a locus that cannot be occupied.
Evidence that argues against this (1)
- The deposit's own paper attributes between-sample occupancy differences primarily to oncoprotein LEVEL and to lineage state rather than to partner identity, and reports that lineage-switching samples have reduced oncoprotein levels - which would produce this pattern without any partner effect. Janssens DH et al. 2024
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The KMT2A N-terminal fragment retained in all six of these fusions is the same protein region in each; the partner supplies the C-terminal half. The two antibodies used here (KMT2A-N1 Millipore 05-764, KMT2A-N2 CST D2M7U) recognise that shared fragment, which is why one antibody pair can be compared across six different fusions at all.
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 · Nat Commun 15:9341 2024What this evidence cannot show
The N/C antibody logic is the deposit paper's own design and is taken from it rather than re-derived. The N antibodies cannot distinguish the fusion from residual wild-type full-length KMT2A, which is why the deposit also carries C-terminal antibodies; this analysis reports N and carries C as a contrast.
Evidence that argues against this (1)
- A shared N-terminal fragment does not guarantee a shared address: the deposit paper's central claim is that fusion-partner-specific binding patterns over lineage gene sets correlate with each mutation's prevalence in ALL versus AML, i.e. that the partner does redistribute the oncoprotein. Janssens DH et al. 2024
the 17 occupied samples at 2.94-29.11x demonstrably carry oncoprotein, so the cohort bound is not a depth artefact
the arm shares occupancy with non-AFF1 samples MORE than chance, which is what a shared HOXA programme would look like
removes the shared-window enrichment as support for fusion targeting without touching the HOXA result itself
does not touch the 17-of-18 positives; qualifies the single negative they rest beside
the same partner-restriction at the transcript level in 48 patients instead of 8 occupancy samples
HOXA is the partner invariant and IRX1 is not; this adds the tier position