The null the claim needed: the KMT2A BCR ranks 72 of 18,082 chromosome-11 bins for its H3K9me1 response - top 0.40% - while the same locus ranks 3,877th for H3K9me3 and 10,276th for H3K9me2
also called top 0.40% of chr11, measured · the rank that makes 'one of the strongest events' checkable · the BCR's percentile for each H3K9 state
Saying a change at one gene is 'one of the strongest in the genome' only means something if you rank it against everywhere else. Ranked against every comparable stretch of chromosome 11, the KMT2A gene is in the top half of one percent for one tag and unremarkable for the other two.
The same thing, for a scientist
The founding paper's locus claim is a superlative - the *KMT2A* breakpoint cluster region was among the strongest H3K9me1 changes across the genome - and a superlative needs a denominator. Chromosome 11 was divided into **18,083 bins of the BCR's own 7,466 bp width**, the two-replicate mean of siKDM3B and control was taken per bin from the deposited tracks, and the log2 ratio computed wherever both arms exceeded a 0.05 floor, leaving **18,082 evaluable bins** [E1]. **H3K9me1: the BCR bin's log2 ratio is +1.160, rank 72, top 0.40%.** That is an independent, quantitative confirmation of the paper's superlative from its own data - and it is a rank on chr11, not the genome, so it is a weaker denominator than the claim it supports. **H3K9me3: +0.397, rank 3,877, top 21.4%** - a modest, unremarkable shift, which is the genome-scale form of this beat's element-level null. **H3K9me2: -0.033, rank 10,276, top 56.8%** - the median bin of chromosome 11, i.e. nothing at all at the BCR, consistent with this beat's finding that the me2 change is confined to the flank. **Three marks, one locus, one pipeline: 0.4th, 21st and 57th percentile.** The separation between the me1 and me3 percentiles is the quantitative statement of mark specificity that the element-level ratios (3.74x versus 1.01x) make locally.
Why a rank and not a p-value
Two replicates per arm cannot support a per-bin significance test, and this graph has published the rule that a proportion is not a result until its null is written down. The rank against all comparable bins is the null, and it is the same framing the paper itself uses for its me1 claim. It has three properties that a p-value here would not: it needs no distributional assumption, it is invariant to the absence of spike-in normalisation as long as the same denominator is used for every bin, and it is directly interpretable - 'the 72nd largest of 18,082' is a sentence a reader can check.
The floor matters and is stated. Bins where either arm's mean fell below 0.05 were excluded, which removed exactly one of 18,083 bins - the last, partial bin. So the denominator is effectively all of chromosome 11 at this bin width and the ranking is not being flattered by dropping low-signal regions. Had the floor been higher, the ranking would improve artificially by removing quiet bins; it was deliberately set low for that reason.
Three things this does not do. (a) It is chromosome 11 only, not genome-wide, so it cannot reproduce a genome-scale superlative - it is the strongest available cheap approximation to one, and it is weaker than the claim. (b) The bin grid is fixed from coordinate 0, so the BCR is not centred in its bin; the bin containing the BCR midpoint (index 15,853) contains most but not all of the BCR, which slightly dilutes it and if anything understates the rank. (c) A 7,466 bp bin cannot separate the BCR from a kilobase of flank, and the flank is where the me2 change lives - so the me2 result here is a statement about the BCR-width bin, not about the mark's behaviour across the locus.
The three percentiles beside the two local ratios
| mark | BCR log2(siKDM3B/control) | rank of 18,082 | percentile | chr11 MEDIAN log2FC | BCR excess over the median | element-level ratio |
|---|---|---|---|---|---|---|
| H3K9me1 | +1.160 | 72 | top 0.40% | +0.004 | +1.156 (2.23x) | 3.74x |
| H3K9me3 | +0.397 | 3,877 | top 21.4% | +0.114 | +0.283 (1.22x) | 1.01x |
| H3K9me2 | -0.033 | 10,276 | top 56.8% | +0.021 | -0.054 (0.96x) | 0.44x |
The median column matters and it was nearly left out. Chromosome 11 as a whole shifts UP for H3K9me3 on KDM3B depletion (median log2FC +0.114, about +8%) while it does not move at all for H3K9me1 (+0.004). So roughly a quarter of the BCR's apparent me3 log-rise is a chromosome-wide shift that has nothing to do with the locus, and the locus-specific part is 1.22x rather than 1.32x. The equivalent correction to H3K9me1 is negligible. A rank alone would have hidden this, because the rank is computed against the same shifted distribution.
The me3 row is the interesting one and it cuts slightly against the tidy reading. At the CTCF element H3K9me3 does not move at all (1.01x). Over a 7,466 bp bin it moves enough to sit in the top fifth of chromosome 11. Those are both true, and together they say the me3 response is a broad, low-amplitude feature of the wider locus rather than a change at the element - which is the same conclusion the balance-signature test reached by a different route, since both single knockdowns raise me3 over the BCR equally. A reader who wanted the cleanest possible mark-specificity story would quote only the 1.01x and only the element; this entry exists so that the 21st percentile is on the record beside it.
And the me2 row is a genuine surprise against the paper's abstract, which describes KDM3B depletion as increasing H3K9me1/2 at the locus. At BCR-width resolution the me2 change at this locus is the median change on chromosome 11 - literally nothing. See mechanism-kdm3b-loss-shifts-the-kmt2a-anchor-from-h3k9me2-to-h3k9me1.
| script | analyses/q108-gse210480-h3k9me3-kmt2a-bcr/run.py |
|---|---|
| status | run |
| run date | 2026-09-22 |
| result summary | the KMT2A BCR bin ranks 72 of 18,082 evaluable chr11 bins for the H3K9me1 siKDM3B/control log ratio (top 0.40%), 3,877th for H3K9me3 (top 21.4%) and 10,276th for H3K9me2 (top 56.8%) |
| bin width bp | 7466 |
| n bins | 18083 |
| n bins evaluable | 18082 |
| signal floor | 0.05 |
| bins excluded by floor | 1 |
| bcr bin index | 15853 |
| why a rank and not a p value | two replicates per arm cannot support a per-bin significance test; a rank against all comparable bins needs no distributional assumption, is invariant to the missing spike-in as long as the denominator is shared, and is directly checkable by a reader. It is also the framing the founding paper itself used for its H3K9me1 locus claim. |
| the tension it records | H3K9me3 does not move at the CTCF element (1.01x) yet sits in the top fifth of chr11 over a BCR-width bin. Both are true. Together they describe a broad, low-amplitude me3 feature of the wider locus rather than a change at the element - the same conclusion the balance-signature test reaches, since both single knockdowns raise me3 over the BCR equally. |
| unrun sensitivity analysis | the rank was computed at one bin width (7,466 bp, the BCR's own). A rank at 2 kb or 20 kb could move the BCR's percentile substantially and was not computed. |
| localization | chromatin across the whole of chromosome 11, with the KMT2A locus at 11q23.3 as the bin of interest |
| subnuclear localization | not measured - a chromosome-wide binned ranking has no subnuclear coordinate |
| cell type | hTERT-immortalised human retinal pigment epithelial (RPE) cells - epithelial, non-haematopoietic, no KMT2A rearrangement |
| cell state | asynchronous proliferating culture 72 h after control or KDM3B siRNA transfection |
| epigenetic readout | H3K9me1, H3K9me2 and H3K9me3 ChIP-over-input coverage, binned at 7,466 bp across chromosome 11 and ranked by siKDM3B/control log2 ratio |
| shown for fusion | Not shown for any fusion. The ranking is over unrearranged RPE chromatin; no fusion-bearing cell contributed a single bin. |
| absence class | measured-not-analysed |
| chromosome wide shift note | chromosome 11's MEDIAN siKDM3B/control log2 ratio is +0.114 for H3K9me3 but only +0.004 for H3K9me1, so about a quarter of the BCR's apparent me3 rise is a chromosome-wide shift. Median-corrected, the locus-specific me3 change is 1.22x against H3K9me1's 2.23x. A rank cannot show this because it is computed against the same shifted distribution. |
inputs
[ "GSE210480 H3K9me1, H3K9me2 and H3K9me3 bigWigs, control and siKDM3B arms, two replicates each, RPE, hg19", "chr11 binned at 7,466 bp - the KMT2A BCR core's own width - across the true hg19 length 135,006,516 bp" ]
ranks — the numbers behind the chart
{
"H3K9me1": {
"log2fc": 1.1597,
"rank": 72,
"percentile_from_top": 0.398,
"chr11_median_log2fc": 0.0044,
"excess_over_median_log2": 1.1553
},
"H3K9me3": {
"log2fc": 0.3969,
"rank": 3877,
"percentile_from_top": 21.441,
"chr11_median_log2fc": 0.1138,
"excess_over_median_log2": 0.2831
},
"H3K9me2": {
"log2fc": -0.0326,
"rank": 10276,
"percentile_from_top": 56.83,
"chr11_median_log2fc": 0.0212,
"excess_over_median_log2": -0.0538
}
}- A rank is persuasive and cheap, and it can launder a weak measurement. The 72-of-18,082 figure rests on two replicates per arm averaged before ranking, with no mappability filter and no spike-in; a single high replicate in the siKDM3B arm at that bin would move it. The rank confirms the direction of the paper's superlative, not its precision.
- The me3 21st-percentile figure is the number in this entry most likely to be quoted out of context - in either direction. It is neither a null (21% is not nothing) nor a positive (a 1.32-fold bin change with n = 2 is not much), and it sits beside an element-level ratio of 1.01x that says the change is not where the mechanism acts.
- Does the BCR's top-0.40% H3K9me1 rank survive a genome-wide denominator? Chromosome 11 is 4.4% of the genome and the locus's own chromosome is not a neutral comparison set - a chromosome-level effect would inflate or deflate every bin on it together. Twenty-three more chromosomes is twenty-three more passes over the same files.
- How does the rank move with bin width? The width was set to the BCR's own, which is the fairest single choice and also the one most likely to flatter this locus. A width sweep from 1 kb to 50 kb would show whether the top-0.40% figure is a property of the locus or of the grid.
- What are the 71 chr11 bins that beat the KMT2A BCR for H3K9me1 response? A control count nobody opens is not a control. If those 71 bins contain other rearrangement-prone genes, that is a finding; if they are pericentromeric or repeat-rich, the ranking needs a mappability filter it does not have.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
Chromosome 11 divided into 18,083 bins of 7,466 bp (the KMT2A BCR core's own width) from the deposited GSE210480 tracks, with two-replicate arm means and a 0.05 signal floor leaving 18,082 evaluable bins, gives the BCR bin a siKDM3B/control log2 ratio of +1.160 and rank 72 (top 0.40%) for H3K9me1, +0.397 and rank 3,877 (top 21.4%) for H3K9me3, and -0.033 and rank 10,276 (top 56.8%) for H3K9me2.
GEO GSE210480 H3K9me1/2/3 bigWigs binned across the true hg19 chr11 length with pybigtools 0.3.0 zoom-level summaries, 2026-09-22; code at analyses/q108-gse210480-h3k9me3-kmt2a-bcr/run.py (section 8) and output in results.json under chr11_null · Gray ZH, Chakraborty D, Duttweiler RR, Alekbaeva GD, Murphy SE, Chetal K, Ji F, Ferman BI, Honer MA, Wang Z, Myers C, Sun R, Kaniskan HU, Toma MM, Bondarenko EA, Santoro JN, Miranda C, Dillingham ME, Tang R, Gozani O, Jin J, Skorski T, Duy C, Lee H, Sadreyev RI, Whetstine JR · NCBI GEO (data accompanying Cell 186:4528-4545.e18) 2023What this evidence cannot show
CHROMOSOME 11 ONLY, not genome-wide, so this approximates rather than reproduces a genome-scale superlative and is a weaker denominator than the claim it supports. The bin grid starts at coordinate 0 so the BCR is not centred in its bin (index 15,853), which dilutes it slightly and if anything understates the rank. A 7,466 bp bin cannot separate the BCR from its flanks, which matters specifically for the H3K9me2 row because the me2 change is flank-confined. Two replicates per arm, averaged before ranking, so a single outlier replicate propagates. No spike-in. The 100 kb HOMER pad at the end of the declared chr11 length was excluded by using the true hg19 length of 135,006,516. RPE cells.
Evidence that argues against this (1)
- A rank is sensitive to the bin width chosen, and the width here was chosen to equal the BCR - which is the fairest choice for THIS locus and also the choice most likely to flatter it, since any other locus whose response happens to be wider or narrower than 7,466 bp is penalised by the same grid. A rank computed at 2 kb or 20 kb bins could move the BCR's percentile substantially, and that sensitivity analysis was not run. Iqbal Lab oncofusion knowledge graph 2026
supplied the chr11 null: me1 rank 72 of 18,082, me3 3,877, me2 10,276
the null that makes the locus claim checkable: BCR rank 72 of 18,082 for me1, 3,877th for me3
independently confirms the H3K9me1 superlative at chromosome scale (top 0.40% of chr11)
H3K9me2 at the BCR is the median chr11 bin - no change at all at BCR-width resolution
a genuine bin-level null, but not matched on baseline mark level - the same correction may apply and is testable from data already read