On an absolute scale, H3K79me2 in the index fusion falls 52-fold (EPZ5676) to 91-fold (compound 10) by day 9 and NOT ONE of 17,341 measurable genes holds its signal - the first absolute-scale chromatin statement this project has been able to make about KMT2A::AFF1
also called nothing is spared in RS4;11 · 52x and 91x, absolutely · zero genes hold at day 9
Using a fruit-fly yardstick added to every sample, we can say for the first time how much of a chromatin mark is actually lost in the leukaemia cells this project is built around, rather than how it is redistributed. Nine days of a DOT1L inhibitor removes about 98 to 99 percent of it everywhere, and of 17,341 genes with enough signal to measure, exactly zero hold on to what they had.
The same thing, for a scientist
**Statement, ABSOLUTE.** In RS4;11 (KMT2A::AFF1, CVCL_0093), restricted to the 19,214 TSS entries that are genuinely RS4;11 in this deposit and to the 17,341 of them with a DMSO day-9 baseline at or above 25: | contrast (day 9) | global fold DOWN, ChIP-Rx | same, depositors' convention | median gene retains | genes holding or gaining | |---|---:|---:|---:|---:| | EPZ5676 1 uM | **52.4x** | 44.0x | **4.71%** | **0 of 17,341** | | compound 10 100 nM | **90.6x** | 60.8x | **1.96%** | **0 of 17,341** | MOLM13 (KMT2A::MLLT3) runs in parallel over 54,958 measurable genes: **63.5x** down under EPZ5676 with a median 3.60% retained, **36.9x** under compound 10 with 4.46% retained, and **0 genes holding in either**. **Pre-registered branch A1 fires, reported as branch A3 requires.** The pre-registration said A1 would fire if under 5% of genes hold. The measured figure is **0.00% in all four line-by-drug contrasts**. But the pre-registration also warned that the two branches both assumed the answer was a count, and that the honest object might be the distribution - so the distribution is what is reported. The 'spared' genes are entirely a low-count floor: at any non-zero baseline 164 RS4;11 genes (0.86%) appear to hold under EPZ5676; at a baseline of 5 that falls to 136; **at a baseline of 25 it is zero and stays zero at 100.** The apparent sparing is counting noise at tiny denominators, and it is reported at four thresholds precisely so a reader can watch it disappear. **Pseudocount is not doing the work.** Of the 17,341 measurable RS4;11 genes, **0** have a treated EPZ5676 day-9 value of exactly zero and **3** (0.02%) under compound 10; in MOLM13, 58 (0.11%) and 20 (0.04%). The `+0.04` pseudocount touches under one gene in a thousand.
Which statements here are ABSOLUTE and which are RELATIVE
Stated explicitly, because this is the distinction this thread has most often got wrong.
ABSOLUTE - depends on the dm6 spike-in and could not be made from any other chromatin dataset in this project:
- the global fold-down figures (52.4x, 90.6x, 63.5x, 36.9x)
- 'the median gene retains 4.71% of its H3K79me2'
- 'zero genes hold or gain'
RELATIVE - would be identical on the raw unscaled table, because absolute = raw x K with K a constant:
- any ranking of genes by loss
- any comparison of one gene class with another
- the fact that the retention distribution is broad
The two scale-factor conventions bound the absolute magnitudes: ChIP-Rx and the depositors' % fly convention differ by at most 1.59x across the twelve contrasts, so the honest reading of '52.4x' is 'between about 44x and 52x'. 12
Why the count of spared genes is reported at four thresholds and not once
There are no replicates in this deposit - one library per condition per timepoint - so there is no measurable noise floor and none may be imported. A gene whose DMSO baseline is a handful of counts can 'hold' its signal purely by sampling. The threshold series is the substitute for an error bar:
| DMSO d9 baseline | n measurable | rows dropped | median retained | holding or gaining |
|---|---|---|---|---|
| any non-zero | 19,142 | 72 | 8.53% | 164 (0.86%) |
| >= 5 | 19,110 | 104 | 8.45% | 136 (0.71%) |
| >= 25 | 17,341 | 1,873 | 4.71% | 0 |
| >= 100 | 9,816 | 9,398 | 1.25% | 0 |
(RS4;11, EPZ5676 day 9.) Every dropped row is counted. The monotone fall of the median with the threshold is itself the tell: the higher the baseline, the more completely the mark is removed - which is the subject of a sibling entry. 1
What this does and does not transfer to
- Not substitutable into cycles 102-107. Those score binned-genome share statistics; this is TSS-averaged sitepro over a -1 kb/+3 kb window on hg19, a different measurement of a different quantity.
- Not a menin-inhibitor statement. The H3K79me2 arm of GSE149183 has no VTP-50469 arm at all. VTP-50469 appears only in the MOLM13 MEN1 and DOT1L occupancy ChIPs (GSM4491617, GSM4491622), which carry no spike-in. Nothing here speaks to menin inhibition, and the standing question of how much of H3K79me2 a menin inhibitor removes on an absolute scale remains unanswerable from any public deposit this project has found.
- An extreme effect size by design. Inhibiting DOT1L abolishes the mark DOT1L writes. That this is visible does not show that spike-in normalisation matters at realistic effect sizes; it shows how far a within-sample share can be from the truth when it does. 13
| script | analyses/absolute-h3k79me2-in-the-index-fusion/run.py |
|---|---|
| status | run |
| result summary | RS4;11 day 9: 52.4x down under EPZ5676, 90.6x under compound 10; median gene retains 4.71% and 1.96%; 0 of 17,341 genes hold. MOLM13: 63.5x and 36.9x; 3.60% and 4.46%; 0 of 54,958 hold. |
| genes holding or gaining baseline ge 25 | 0 |
| n measurable rs411 | 17341 |
| n measurable molm13 | 54958 |
| prereg branch | A1, reported in the form branch A3 required |
inputs
[ "GSE149183 RS4_11 and MOLM13 sitepro tables, md5-verified 2026-09-23" ]
absolute fold down day9 — the numbers behind the chart
{
"RS4;11_EPZ5676_1uM": 52.4,
"RS4;11_compound10_100nM": 90.6,
"MOLM13_EPZ5676_1uM": 63.5,
"MOLM13_compound10_100nM": 36.9
}median PCT retained day9 baseline ge 25 — the numbers behind the chart
{
"RS4;11_EPZ5676_1uM": 4.71,
"RS4;11_compound10_100nM": 1.96,
"MOLM13_EPZ5676_1uM": 3.6,
"MOLM13_compound10_100nM": 4.46
}- Near-total removal of H3K79me2 is achievable and was achieved clinically by pinometostat, and it still produced only modest responses. A larger measured chromatin effect is therefore not evidence of a better drug - it is evidence that this pharmacodynamic endpoint saturates well before the clinical one does.
- If a DOT1L inhibitor removes 98-99% of H3K79me2 genome-wide, then essentially every gene in a normal haematopoietic cell loses it too. The therapeutic index cannot come from selectivity of the chromatin effect; it must come from differential dependence on the mark.
- A compound that drives the fly fraction of the library to 34% is removing so much human signal that library complexity itself becomes the limiting factor for any downstream measurement - which is a practical limit on using H3K79me2 ChIP as a pharmacodynamic assay at deep inhibition.
- What is the absolute figure for a MENIN inhibitor? No deposit found by this project carries a spike-in-normalised histone ChIP under VTP-50469, revumenib or ziftomenib, so the single most decision-relevant number in this thread is still unmeasured.
- Is the residual 1-5% real H3K79me2 or antibody background? With no IgG or no-primary control in the spike-in series, the floor of the assay is unknown, and it bounds every retention figure from below.
- H3K79me2 has no known demethylase in this graph's reading of the literature, so a 52-91x fall in nine days implies dilution by replication plus turnover. Does the measured rate match the line's doubling time (RS4;11 ~35-60 h)? That arithmetic is one afternoon and would test whether passive dilution alone can account for it.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
With the deposited dm6 counts applied as a ChIP-Rx scale factor, H3K79me2 at TSS windows in RS4;11 falls 52.4-fold under EPZ5676 and 90.6-fold under compound 10 by day 9, and no gene with a DMSO baseline at or above 25 holds its signal.
GSE149183 series supplementary files: RS4_11 and MOLM13 H3K79me2 TSS-ChIPseq signal + dm6 spike-in sitepro tables · Perner F, Gadrey JY, Xiong Y, et al. · NCBI Gene Expression Omnibus 2020What this evidence cannot show
Confidence is 'strong', not 'established', for one reason: there is ONE library per condition, so the point estimate has no sampling distribution. The effect is far larger than any noise floor this project has measured (2-16%), which is why the direction and rough magnitude are safe; the third significant figure is not. Restricted to 19,214 of 57,794 TSS entries because two-thirds of the RS4;11 file is MOLM13, and that retained subset has a median baseline of 107.6 against 53.6 for the full annotation, so it is signal-enriched and not a random sample.
Evidence that argues against this (1)
- If Drosophila chromatin is immunoprecipitated less efficiently once the human H3K79me2 pool collapses - a real possibility for an antibody-based spike-in, since the fly fraction of the library rises to 34% - then the fly counts overstate the constant input and the true fold-down lies between the raw 7.4x and the scaled 52.4x. Nothing in this deposit resolves it. The ChIP-Rx method paper is explicit that the assumption of constant spike-in recovery is an assumption. Orlando DA et al. 2014
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
EPZ-5676 (pinometostat) is a selective DOT1L methyltransferase inhibitor that depletes H3K79 methylation, and the Blood 2020 study reports compound 10 as a more potent novel DOT1L inhibitor tested at 100 nM against EPZ5676 at 1 uM.
Novel inhibitors of the histone methyltransferase DOT1L show potent antileukemic activity in patient-derived xenografts. · Perner F, Gadrey JY, Xiong Y, Hatton C, Eschle BK, Weiss A, et al. · Blood 2020What this evidence cannot show
The two drugs are given at different concentrations (1 uM vs 100 nM), so any drug-to-drug comparison here confounds compound identity with dose. Only within-drug comparisons are clean.
Evidence that argues against this (1)
- Pinometostat's clinical results in KMT2A-rearranged leukaemia were modest despite near-complete H3K79me2 depletion, so a large chromatin effect measured here does not predict a large clinical effect. This graph already records that as the most instructive failure in the field. Stein EM et al. 2018
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
RS4;11 (CVCL_0093) carries an in-frame KMT2A::AFF1 fusion and is this project's index fusion.
Cellosaurus record CVCL_0093 (RS4;11) · Bairoch A · Cellosaurus, SIB 2026What this evidence cannot show
Fetched from the Cellosaurus REST API on 2026-09-23. The line is an adult (32Y, female) B-ALL line; the index fusion's clinical centre of gravity is infant ALL, so the cell context is not the patient context.
Evidence that argues against this (1)
- A cell line is not a patient. RS4;11 also carries an NSD2 p.Glu1099Lys mutation, which is itself a chromatin lesion and could modulate H3K79me2 dynamics independently of the fusion. Cellosaurus, SIB 2026
- related to →KMT2A::AFF1 (MLL-AF4), t(4;11)(q21;q23)
the first absolute-scale chromatin measurement this project holds for the index fusion
98-99% of the mark is removed at every measurable TSS by day 9
- tests →Pinometostat (EPZ-5676) — the DOT1L inhibitor, and the most instructive failure in this field
EPZ5676 1 uM, day 9, spike-in scaled: 52.4x global loss in RS4;11, 63.5x in MOLM13
inhibiting DOT1L removes its product essentially completely and uniformly
the absolute scale used here is applied by this beat, because the deposited columns are raw
restricted to the 19,214 rows that are genuinely RS4;11
confirmed independently: the spike-in H3K79me2 series is DMSO / EPZ5676 / compound 10 only
a 52-91x fall over nine days is the quantity a dilution-plus-turnover model has to account for
three timepoints on an absolute scale is the closest this corpus has come to a decay rate, and it is still not a half-life
the per-gene retention figures come from the same scaled table
the day-9 endpoint of the same scaled series
what survives the design: effect sizes two orders of magnitude above any plausible floor
bounds the retention figures from below; the fold-down is a floor on the effect, not a ceiling