start / graph / ana-the-marks-respond-proportionally-at-the-peaks-and-quantitatively-in-the-bulk-and-h3k79me2-reverses-sign

Analysismoderate

The first measurement of any histone mark outside the top 0.1% of fusion sites: by ratio the drug's mark response is confined to the peaks as pre-registered, while in ABSOLUTE share the movement at the decile below is 5 to 33 times larger - though that is partly guaranteed by the stratum being 100x bigger. H3K79me2 falls at the peaks (0.8065) and is not measurably changed at D1, and H3K4me3 and H3K79me2 both break the floor at the lowest decile

also called the marks outside S_FLAG, measured for the first time · proportionally confined, quantitatively not · H2AK119ub is unscoreable at D1

Two earlier studies here showed a drug changing four chemical marks on DNA packaging, but both only looked at the strongest one-in-a-thousand sites, and this is the first time anyone has looked anywhere else. Measured as percentages the changes really are concentrated at those strongest sites, which supports what was done before, while in sheer quantity far more movement happens in the surrounding region - though that is partly just because the surrounding region is a hundred times bigger.

The same thing, for a scientist

**Statement.** Q125 and Q130 measured four marks across Q117's eviction quartiles, which are cut **inside** `S_FLAG` — the top 0.1% of fusion bins, 1,174 bins holding 0.84% of the FLAG signal. Cycle 103 found the drug's effect on fusion occupancy is genuinely confined there, and closed by naming the gap: **no mark had ever been measured outside it.** `D1` alone is **117,314 bins holding 19.4%**. Measured. **VTP/DMSO by stratum, control genotype:** | mark | TOP | **D1** | D2 | D5 | **D10** | |---|---:|---:|---:|---:|---:| | H2AK119ub `PAIRED` | 1.1255 | 1.0524 | 1.0065 | 0.9822 | 0.9846 | | H2AK119ub `REP2` | 0.9387 | 0.9784 | 1.0071 | 1.0076 | 1.0027 | | H3K27me3 | 1.2798 | 1.0478 | 1.0254 | 0.9987 | **0.9375** | | H3K4me3 | 1.1235 | 1.0982 | 0.9324 | 0.9018 | **0.8744** | | **H3K79me2** | **0.8065** | **1.0777** | 1.0220 | 0.9386 | **0.8383** | **`M1` fires as pre-registered:** no scoreable mark's **ratio** at D1 departs from 1 by more than the 10% floor. Proportionally, the chromatin response is a peak phenomenon and the thread's `S_FLAG` focus was right. **And the pre-registration measured the wrong half, which its own `U3` clause anticipated:** *"a ratio near 1 in a stratum 100× larger is not the same as no effect."* In **absolute share change**, D1 dwarfs the top stratum: | mark | TOP | D1 | **|D1| / |TOP|** | |---|---:|---:|---:| | H2AK119ub `PAIRED` | +0.000279 | +0.009251 | **33.1×** | | H3K27me3 | +0.000204 | +0.005473 | **26.8×** | | H2AK119ub `REP2` | −0.000176 | −0.004244 | 24.1× | | H3K4me3 | +0.004484 | +0.044805 | 10.0× | | H3K79me2 | −0.006284 | **+0.032526** | 5.2× | **So the drug's mark response is proportionally concentrated at the peaks and quantitatively concentrated in the bulk.** Both statements are true of the same data and the thread has only ever seen the first. **H3K79me2 falls at the peaks and is not measurably changed in the bulk.** It is **0.8065** at TOP (−0.0063 of share) and **1.0777** at D1 (+0.0325). **I first wrote that as 'reverses sign', and this entry's own counter-evidence retracts it:** the D1 value sits *inside* the borrowed floor, so by this entry's own standard only the peak fall is established. The absolute movement at D1 is five times larger and its direction is not distinguishable from noise. Q130's headline — the eviction-depth gradient in H3K79me2, span 0.6005 — was measured entirely inside the stratum where the mark goes down, while five times more of its absolute movement is a gain outside it. **Two marks break the floor at the lowest decile:** H3K4me3 **0.8744** and H3K79me2 **0.8383** at D10, both beyond 10%. **The pre-registration asked only about D1**, because that is the stratum cycle 103 named. It fires clean there and misses this.

Go deeper

H2AK119ub is unscoreable at D1, exactly as the clause predicted

U2 was written because Q125 had found the two H2AK119ub conventions disagreeing in sign on an aggregate contrast. At D1 they disagree again: PAIRED gives +0.0524 and REP2 gives −0.0216. One says the mark rises in the bulk under drug, the other says it falls.

So H2AK119ub is UNSCOREABLE at D1 and is excluded from the verdict rather than averaged. Averaging them would have produced +0.015, a small positive number that looks like a measurement and is the midpoint of a contradiction.

That is the second beat in which these two conventions have disagreed, on different statistics. The mark has a reproducibility problem in this deposit that is independent of which question is asked of it, and any future use of H2AK119ub from GSE260742 should report both conventions or neither.

What the control can and cannot do here, stated because it is weaker than it looks

Both input pairs were stratified identically, and cycle 102 used them to derive the 4–10% floor this verdict leans on. But the inputs have no drug arm. They bound genotype variation — how much two cell lines differ in coverage — and this cycle is asking about drug variation within one genotype. That is a different quantity and the inputs do not bound it.

I have used the genotype floor as the scale anyway, because it is the only empirical scale this deposit offers, and labelled it as a substitute rather than a match. If drug-arm variation is wider than genotype variation, M1 may be firing on a floor that is too generous, and the D1 ratios of 1.05–1.10 could be noise rather than the small real effects this entry treats them as.

The fix is the same one that three other questions in this thread are now blocked on: a same-condition replicate pair, which GSE260742 does not contain and GSE299100 does.

Facts and figures

question	do the histone marks move anywhere except the top 0.1% of fusion sites?
answer	by RATIO at D1, no - M1 fires as pre-registered. By ABSOLUTE share change, 5-33x more movement happens at D1 than at the peaks. Both are true.
downloads required	0
first measurement of any mark outside S FLAG	true
H3K79me2 reverses sign	falls at the peaks (0.8065, -0.0063 of share) and rises at D1 (1.0777, +0.0325). Stated as a direction only - the D1 arm sits inside the borrowed floor and is not distinguishable from noise by this entry's own standard.
the floor is borrowed from the wrong quantity	cycle 102's 4-10% bounds GENOTYPE variation from input pairs with no drug arm; this cycle asks about DRUG variation within one genotype. Used as the only empirical scale available and labelled a substitute.
absence class	measured-not-analysed
absence class note	Every mark track and both input pairs were binned and on disk before this cycle; what had never been analysed is any mark outside the top 0.1%, because every beat in this thread stratified inside S_FLAG.

VTP over DMSO NT

absolute share change NT

stratum sizes

{
  "TOP": 1174,
  "each_decile": 117314
}

VTP over DMSO NT — the numbers behind the chart

{
  "H2AK119ub_PAIRED": {
    "TOP": 1.1255,
    "D1": 1.0524,
    "D2": 1.0065,
    "D5": 0.9822,
    "D10": 0.9846
  },
  "H2AK119ub_REP2": {
    "TOP": 0.9387,
    "D1": 0.9784,
    "D2": 1.0071,
    "D5": 1.0076,
    "D10": 1.0027
  },
  "H3K27me3": {
    "TOP": 1.2798,
    "D1": 1.0478,
    "D2": 1.0254,
    "D5": 0.9987,
    "D10": 0.9375
  },
  "H3K4me3": {
    "TOP": 1.1235,
    "D1": 1.0982,
    "D2": 0.9324,
    "D5": 0.9018,
    "D10": 0.8744
  },
  "H3K79me2": {
    "TOP": 0.8065,
    "D1": 1.0777,
    "D2": 1.022,
    "D5": 0.9386,
    "D10": 0.8383
  }
}

absolute share change NT — the numbers behind the chart

{
  "H2AK119ub_PAIRED": {
    "TOP": 0.000279,
    "D1": 0.009251,
    "ratio": 33.1
  },
  "H3K27me3": {
    "TOP": 0.000204,
    "D1": 0.005473,
    "ratio": 26.8
  },
  "H2AK119ub_REP2": {
    "TOP": -0.000176,
    "D1": -0.004244,
    "ratio": 24.1
  },
  "H3K4me3": {
    "TOP": 0.004484,
    "D1": 0.044805,
    "ratio": 10
  },
  "H3K79me2": {
    "TOP": -0.006284,
    "D1": 0.032526,
    "ratio": 5.2
  }
}

two marks break the floor at D10

{
  "H3K4me3": 0.8744,
  "H3K79me2": 0.8383,
  "note": "the pre-registration asked only about D1, because that is the stratum cycle 103 named"
}

U2 fired

{
  "paired_minus_1": 0.0524,
  "rep2_minus_1": -0.0216,
  "same_sign": false,
  "consequence": "H2AK119ub is UNSCOREABLE at D1 and excluded from the verdict rather than averaged; averaging would give +0.015, the midpoint of a contradiction",
  "second_occurrence": "the two conventions also disagreed on an aggregate contrast in Q125, so the mark has a reproducibility problem in this deposit independent of the question asked"
}

Still open

Is |D1|/|TOP| informative at all? D1 has 100x the bins, so identical proportional behaviour would put far more absolute change there. The test that separates them is whether D1's change exceeds what its size predicts - which is the ratio, and the ratio says no. This entry gives the absolute framing equal weight and may be overstating it.
Does H3K79me2's D1 rise survive a drug-arm floor? It sits inside the borrowed genotype floor, so by this entry's own standard only the peak fall is established. A same-condition replicate would settle it - GSE299100 has one and GSE260742 does not.
Why do two marks break the floor at D10 and not at D1? H3K4me3 0.8744 and H3K79me2 0.8383 in the lowest-signal decile, where signal is sparsest and ratios least stable. That is either a real depletion in the bulk or the noisiest stratum behaving noisily, and the input pairs at D10 (1.0256 and 0.9931) mildly favour the former.
Should H2AK119ub from GSE260742 be used at all? Its two conventions have now disagreed in sign on two different statistics in two different beats. Any entry quoting it should carry both or neither.

Every source behind this entry (1)

E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.analysis
Over the parental FLAG DMSO universe cut into the top 0.1% (1,174 bins) plus deciles (117,314 bins each), VTP/DMSO share ratios in the control genotype are: H2AK119ub PAIRED 1.1255 at TOP and 1.0524 at D1; H2AK119ub REP2 0.9387 and 0.9784; H3K27me3 1.2798 and 1.0478; H3K4me3 1.1235 and 1.0982; H3K79me2 0.8065 and 1.0777. At D10 the ratios are 0.9846, 1.0027, 0.9375, 0.8744 and 0.8383 respectively. Absolute share changes at D1 exceed those at TOP by factors of 33.1 (H2AK119ub PAIRED), 26.8 (H3K27me3), 24.1 (H2AK119ub REP2), 10.0 (H3K4me3) and 5.2 (H3K79me2). The two H2AK119ub conventions differ in sign at D1, +0.0524 against -0.0216.
analyses/do-the-marks-move-in-the-bulk/run.py and results.json, this repository; bins from C:/Temp/q117, C:/Temp/q124 and C:/Temp/q125, computed 2026-09-23 with zero downloads; the floor is from cycle 102's two input pairs · Oncofusion knowledge graph, coordinator · internal analysis, reproducible from the repository 2026
What this evidence cannot show
THE FLOOR IS BORROWED FROM THE WRONG QUANTITY. Cycle 102's 4-10% bounds GENOTYPE variation from two input pairs; this cycle asks about DRUG variation within one genotype, and the inputs have no drug arm. If drug-arm variation is wider, M1 fires on a floor that is too generous and the D1 ratios of 1.05-1.10 may be noise. WITHIN-SAMPLE SHARE ONLY - no spike-in exists in GSE260742, so a uniform global change in any mark is invisible and every ratio is relative. n = 1 for H3K27me3, H3K4me3 and H3K79me2 per genotype per treatment; only H2AK119ub has two conventions and they disagree. Strata are defined by PARENTAL FLAG signal from a third experiment, so a bin's stratum reflects where the fusion sits in the parental line, not where the mark sits. The pre-registration asked about D1 specifically, because cycle 103 named it; it does not cover D10, where two marks exceed the floor.
Evidence that argues against this (3)
- The absolute-change comparison is close to guaranteed by the stratum sizes and may carry no information at all. D1 has 100 times the bins and 23 times the signal of the top stratum, so even identical proportional behaviour would put far more absolute change there. The honest test of whether the bulk matters is not |D1|/|TOP| but whether D1's change exceeds what its size alone predicts - which is exactly the ratio, and the ratio says no. This entry presents both and gives the absolute framing equal weight, which may overstate it.
- H3K79me2's sign reversal is the most striking claim here and it has the weakest support. It is one track per condition, no replicate, and the D1 ratio of 1.0777 sits inside the borrowed floor - so by this entry's own standard that arm of the reversal is not distinguishable from noise. Only the TOP value of 0.8065 is clearly outside it. 'Reverses sign' would be better stated as 'falls at the peaks and is not measurably changed in the bulk', which is less interesting and better supported.
- Q130's H3K79me2 result is not undermined by this and the entry could be read as implying it is. Q130 measured a gradient ACROSS eviction quartiles within S_FLAG and controlled it against a matched null, a replicate pair and a human independent deposit. Finding that more absolute mark change happens outside S_FLAG says nothing about whether that gradient is real; it says the stratum is small. The two results are about different things and putting them in the same paragraph invites a conflation the data does not support.

Connected to

supports →PCGF1 loss flattens the fusion's genomic distribution - the knockout-over-control share ratio rises monotonically from 0.810 at the highest-signal decile to 1.136 at the lowest while both input pairs stay flat - and the drug does something spatially different: VTP-50469 evicts almost only at the top 0.1% (0.7554 against 0.9937 at the decile below). The drug acts at the peaks, the genotype acts in the bulk
answers that entry's closing gap - the first measurement of any mark outside the top 0.1%
related to →All four murine VTP-50469 H3K79me2 contrasts in GSE260742 give Spearman -1.000 against eviction depth, and all four SGC0946 contrasts are flat: the sign reproduces four times out of four, the span only to within a factor of 1.99, and the replicate pair that establishes this was already binned on disk from cycle 82
Q130s H3K79me2 gradient was measured entirely inside the stratum where the mark falls; this does not undermine it, the two are about different things
related to →DOT1L — the target where excellent chemistry met an indirect interaction
H3K79me2 falls at the peaks and is not measurably changed in the bulk
related to →KMT2A / MLL1 — the gene that breaks
strata are defined by parental KMT2A-fusion occupancy
← supportsAuditing all 2,476 committed entries for single-convention H2AK119ub: 122 mention the mark, 10 carry both conventions, 2 name the one they use, and 2 quoted a paired-convention number without saying so. Both are amended - and the more useful result is that the two conventions disagree PER STATISTIC rather than globally, so declaring the mark unusable would have been wrong
answers that entry's open question - audits which committed entries relied on one convention, and finds the disagreement is per-statistic rather than global
← related toEnumerating every H2AK119ub number in 168 analysis result files instead of hand-picking fourteen: the two conventions flip sign on 35 of 68 comparable ratios but MATERIALLY on only 2, the exhaustive list finds exactly one real single-convention quotation against two arithmetic coincidences, and the statistic cycle 105 cited as proof they agree is one of the two places they materially disagree
that beat excluded H2AK119ub at D1 as unscoreable on a sign disagreement; this one shows sign flips are common and usually trivial, so exclusion there needs the materiality test rather than the sign alone
← refutesAnswering a question three cycles filed and none ran: H3K79me2 also has two channels in GSE260742 - rep1 and rep2, 199 numeric leaves each - so cycle 104 quoted one of two available numbers without saying so. Unlike H2AK119ub they never materially disagree, 0 of 18 comparable ratios, and H3K27me3 and H3K4me3 were never scored under any second channel at all
that panel quoted H3K79me2 as a single number when two channels exist; the choice was unstated, and the unquoted channel agrees - a provenance fault, not a wrong number
← refutes[CORRECTED cycle 110 - F2 not F1, and the floor spans ~20x across arms] Measuring the same-condition histone-mark noise floor from replicates already binned on local disk, zero downloads: 15.97% at the peaks and 3.03% at D1, so the borrowed 10% floor was 3.3x too generous in the bulk - independently reproducing, in a different deposit, cell system, mark and antibody, what the GSE299100 beat measured at 9.4% and 2.4%
its M1 excluded effects up to 10% at D1 where the measured same-condition floor is 3.03%, 3.3x tighter
← refutesScoring cycle 104's mark panel against a floor measured in its OWN arm: 9 of 10 mark-by-arm cells at D1 land inside the floor's own spread, so D1 is not resolvable from this deposit at all - and the same mark ratios give 0, 4 or 4 breaking marks depending only on which floor was imported, so cycle 104's pass, cycle 107's F1 and the GSE299100 beat's failure were all measuring the ruler rather than the marks
its M1 pass at D1 used a floor 1.5x wider than its own arm's median and 3.3x wider than the other arm's; at D1 nothing in this deposit is resolvable, and at D10 M1 fails in both arms
← refutesCycle 104's M1 - 'no scoreable mark moves at D1, so the menin inhibitor's chromatin effect is proportionally a peak phenomenon' - DOES NOT SURVIVE. The same-condition floor at D1 is 2.4%, not the 10% borrowed from the top stratum, and all three scoreable marks exceed it (H3K4me3 9.8%, H3K79me2 7.8%, H3K27me3 4.8%). Two of the three exceed even GSE260742's OWN stratum-matched input floor, which was on disk the whole time. Cycle 104's alternative M2 fires instead
← refutesCycle 104's open question - 'why do two marks break the floor at D10 and not at D1?' - has a false premise. Against the measured same-condition floor of 2.3% at D10, THREE marks break it there (H3K79me2 16.2%, H3K4me3 12.6%, H3K27me3 6.3%), and the same three break it at D1 too. There is no D1-versus-D10 contrast to explain; there is a continuous bulk gradient that a single borrowed 10% figure had chopped in two