Oncofusion

start / graph / ana-q130-the-dot1l-inhibitor-removes-86-percent-of-h3k79me2-at-these-bins-with-no-eviction-depth-structure-at-all

Analysismoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q130-does-the-eviction-gradient-exist-in-h3k79me2

SGC0946 removes 86% of H3K79me2's share at the fusion's own sites and the four eviction quartiles are 0.1396 / 0.1453 / 0.1452 / 0.1360 - span 0.0036, Spearman +0.400, a bootstrap interval spanning zero - so a very large effect on this mark can carry no eviction-depth structure whatever

also called the flat control · SGC0946 is huge and shapeless · the positive control that proves the gradient is not a generic drug effect

A drug that blocks the enzyme making the tag removes almost all of it everywhere, evenly. The menin inhibitor removes far less of it overall but removes it in a strict pattern that tracks how much fusion protein was pulled off. Size and shape are different measurements, and only the shape says which drug was given.

The same thing, for a scientist

**Same 1,228 bins, same quartiles, same culture, same normalisation, same script - only the drug changes.** In murine MLL-AF9 (KMT2A::MLLT3) NTvKO control, 5 uM SGC0946 for 72 h against DMSO 72 h [E1]: | drug | Q1 | Q2 | Q3 | Q4 | Spearman | span | aggregate | |---|---:|---:|---:|---:|---:|---:|---:| | SGC0946 (DOT1L inhibitor) | 0.1396 | 0.1453 | 0.1452 | 0.1360 | +0.400 | **0.0036** | 0.1415 | | VTP-50469 (menin inhibitor) | 0.5860 | 0.8188 | 0.9280 | 1.1865 | **-1.000** | **-0.6005** | 0.8156 | **The DOT1L inhibitor's effect is five times larger and has no shape.** Its aggregate says the mark's share at the fusion's sites falls to 14% of vehicle; the largest difference between any two quartiles is 0.0093 and Q1-minus-Q4 is 0.0036, against 0.6005 for the menin inhibitor [E1]. **This is the control the gradient result most needed and it was free.** It refutes, in the same bins and the same mark, the reading that *any* drug that moves H3K79me2 will produce an eviction-depth gradient because the quartiles differ in baseline. They differ in baseline under SGC0946 exactly as they do under VTP - the DMSO shares are the same numbers, 0.013029 / 0.008957 / 0.006825 / 0.005918 - and SGC0946 returns a flat profile [E1][E2]. **The per-bin statistics agree.** Under SGC0946 the median per-bin log2 share ratio is -2.8447 / -2.7107 / -2.6805 / -2.7389 - the median bin loses 86.1% of its share and the ordering is not even monotone (Spearman -0.400) - and the bin-level bootstrap 95% interval for the span is [-0.0078, +0.0138], **spanning zero**, in 2,000 resamples of which only 73.1% keep the observed sign. Under VTP the same interval is [-0.651, -0.545] with 100% sign agreement [E3]. **And the matched null shows the SGC effect is genuinely bigger at S_FLAG than elsewhere**, so this is not a flat-because-nothing-happened result: the DMSO-decile-matched bins outside `S_FLAG` give 0.2489 / 0.2493 / 0.2483 / 0.2586, so the fusion's own sites lose about 10 percentage points more share than matched chromatin does - uniformly across all four quartiles [E2].

Go deeper
Why a flat profile is not the same as no effect, and why that distinction is the point

A within-sample share is blind to uniform change (PMID 27875550). So a flat profile could in principle mean 'nothing happened' - but here it cannot, because the aggregate is 0.1415. The mark's share at these bins collapses to a seventh of vehicle and the collapse is the same size in the quarter the fusion was almost entirely evicted from and in the quarter it was barely touched.

That is the correct shape for a writer-enzyme inhibitor: DOT1L is blocked everywhere, so its mark is depleted everywhere it was being deposited, and nothing about that depends on whether the fusion is still present at a given site over 72 hours. It is the wrong shape for the menin inhibitor's effect, which is what makes the two separable.

What this says to cycle 83's threat, precisely and with its limits

Cycle 83 recorded a real problem: a DOT1L inhibitor moves the MLL1-N occupancy channel as much as a menin inhibitor does - Q122's C7 measured 0.3012 under SGC0946 against 0.3298 under VTP-50469 at MLL1's own top sites, both missing a pre-registered floor of 0.85 by about 0.55 - and concluded that this corpus has no validated occupancy-based pharmacodynamic marker for menin inhibition.

Two things must be kept separate. C7's channel is MLL1-N occupancy; this entry's channel is the H3K79me2 histone mark. They are different measurements in different bin sets, and nothing here refutes C7 or its arithmetic.

What this entry adds is that on the H3K79me2 channel, stratified, the two drugs are not confusable. Spearman -1.000 with span 0.6005 against Spearman +0.400 with span 0.0036 and a bootstrap interval spanning zero. The discriminating quantity is the SHAPE of the response across eviction depth, not its magnitude - and on magnitude alone the DOT1L inhibitor's effect is the larger of the two, which is exactly the trap C7 fell into.

The limit, stated plainly: measuring the shape requires knowing each bin's eviction depth, which requires a paired fusion-occupancy ChIP in the same sample. So this is a candidate pharmacodynamic statistic, demonstrated in a murine cell line with a tagged over-expressed construct, and it is not yet a candidate assay for a patient sample.

Facts and figures
scriptanalyses/q130-does-the-eviction-gradient-exist-in-h3k79me2/run.py
statusrun
result summarySGC0946 at S_FLAG: 0.1396 / 0.1453 / 0.1452 / 0.1360, aggregate 0.1415, span 0.0036, bootstrap span CI [-0.0078, +0.0138] spanning zero. VTP-50469: span 0.6005, CI [-0.651, -0.545].
absence classmeasured-not-analysed
absence class noteThe SGC0946 arm of the NTvKO culture was binned by Q124 on 2026-09-23 and scored only as an aggregate (G_NT = 0.1415). Its per-quartile profile - the control that makes the VTP gradient interpretable - was one line of the same script and had never been run.
absence class assigned2026-09-23 (Q130)
epigenetic readoutH3K79me2 within-sample share ratio under a DOT1L inhibitor, eviction-depth stratified, as a shape control for the menin inhibitor
inputs
[
  "GSM8123438 DMSO",
  "GSM8123439 SGC0946 5 uM 72 h",
  "GSM8123440 VTP-50469 500 nM 48 h",
  "GSM8123359/GSM8123361 FLAG (Q117, define the quartiles)"
]
sgc by quartile
[
  0.1396,
  0.1453,
  0.1452,
  0.136
]
vtp by quartile
[
  0.586,
  0.8188,
  0.928,
  1.1865
]
sgc matched null
[
  0.2489,
  0.2493,
  0.2483,
  0.2586
]
sgc residual
[
  -0.1093,
  -0.104,
  -0.1031,
  -0.1226
]
sgc median per bin log2
[
  -2.8447,
  -2.7107,
  -2.6805,
  -2.7389
]
sgc bootstrap span ci95
[
  -0.0078,
  0.0138
]
vtp bootstrap span ci95
[
  -0.651,
  -0.5446
]
dmso share by quartile shared by both contrasts
[
  0.013029,
  0.008957,
  0.006825,
  0.005918
]
How this could bite us back
  • Proposing the eviction-depth SHAPE as a pharmacodynamic readout means proposing a paired fusion-occupancy ChIP in the same patient sample, because the strata cannot be defined without it. In a marrow aspirate with limited cells and no epitope tag that is a serious and possibly fatal practical objection, and this entry does not solve it.
  • If a DOT1L inhibitor removes 86% of H3K79me2's share at the fusion's sites and a menin inhibitor removes 41% at the most-evicted quarter, then adding a DOT1L inhibitor to menin inhibition adds most where the menin inhibitor adds least - at Q3 and Q4. That is an argument for combination, and it is also an argument that the combination's toxicity will be concentrated in normal cells where no eviction gradient exists at all, because DOT1L's function is not fusion-specific.
  • SGC0946 at 5 uM for 72 hours is a tool-compound exposure. Reading its flat profile as 'what a DOT1L inhibitor does' generalises from one compound at one supra-pharmacological concentration in one murine line.
Still open
  1. Does pinometostat, the clinical DOT1L inhibitor, also give a flat profile? No pinometostat arm exists in GSE260742 and no deposited series with a pinometostat H3K79me2 ChIP and a matched fusion-occupancy measurement has been identified.
  2. Is the ten-point excess depletion at S_FLAG under SGC0946 a turnover effect? DOT1L inhibition depletes H3K79me2 by dilution through replication, so the mark should fall fastest where it turns over fastest - which is testable against a replication-timing track and was not attempted here.
  3. What happens under both drugs together? GSE260742 deposits no combination arm, and the combination is the clinically interesting question because the two drugs act on opposite ends of the same eviction axis.
Every source behind this entry (4)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    In murine MLL-AF9 (GSE260742 NTvKO control), 5 uM SGC0946 72 h against DMSO 72 h gives H3K79me2 within-sample share ratios of 0.1396 / 0.1453 / 0.1452 / 0.1360 across the four eviction-depth quartiles of S_FLAG - aggregate 0.1415, span 0.0036, Spearman +0.400 - against 0.5860 / 0.8188 / 0.9280 / 1.1865 for 500 nM VTP-50469 48 h in the same bins.

    What this evidence cannot show

    NO SPIKE-IN NORMALISATION IS POSSIBLE. Q117 verified zero hits for the entire spike-in vocabulary across all 419,824 characters of GSE260742's 158 sample records and no scale factor is deposited, so every cross-condition number here is a WITHIN-SAMPLE SHARE. It cancels the unknown per-sample scale factor exactly and IS BLIND TO A UNIFORM CHANGE BY CONSTRUCTION: nothing here measures H3K79me2's abundance, everything measures its distribution. EXPOSURE ASYMMETRY: DMSO 72 h and SGC0946 5 uM 72 h are matched, VTP-50469 500 nM is 48 h, so every VTP ratio is a LOWER BOUND on the 72-hour effect - identical in both genotypes, so the genotype DIFFERENCE is unaffected. REPLICATION, CORRECTED MID-BEAT: the NTvKO arm deposits ONE H3K79me2 ChIP per genotype per drug, so ITS spans carry no technical error bar and its bin-level bootstrap intervals measure sampling within one library, NOT library-to-library variation. A REPLICATE PAIR DOES EXIST for a different culture - GSM8123372/374 and GSM8123386/388, both titled MLLAF9 Control, the second carrying `rep2` - already binned by Q122 and scored in Q130 DEVIATION 4: the SIGN and the monotonicity reproduce (Spearman -1.000 in all four murine VTP contrasts in this deposit) while the SPAN reproduces only to within a factor of 1.99 (-0.2438 against -0.4847). EVERY SPAN IN THIS BEAT SHOULD BE READ WITH THAT 2-FOLD REPRODUCIBILITY. THE KNOCKOUT is one CRISPR guide, one clone C1, no second clone, no rescue, and NO deposited PCGF1 protein or transcript measurement. S_FLAG is defined by a DYKDDDDK-tag ChIP on a RETROVIRALLY OVER-EXPRESSED tagged MLLAF9 construct, so it is the over-expressed molecule's sites and not an endogenous fusion's. THE H3K79me2 ANTIBODY IS NOT RESOLVED - `antibod` occurs zero times in the 158 GEO records - and is recorded unresolved rather than guessed. SYSTEM: murine MLL-AF9 (KMT2A::MLLT3) myeloid blood cells, NOT the infant B-lineage KMT2A::AFF1 disease this graph is about. EXPOSURE NOTE SPECIFIC TO THIS COMPARISON: SGC0946 is exposure-matched to its vehicle at 72 h while VTP is 48 h, so if anything the VTP arm is UNDER-exposed relative to the SGC arm and its gradient is a lower bound. The comparison of SHAPES is unaffected by an overall exposure difference; the comparison of MAGNITUDES is not, and no magnitude claim is made beyond 'SGC's is larger', which the exposure asymmetry can only strengthen.

    Evidence that argues against this (2)
    • A published methods study establishes that STANDARD ChIP-seq normalisation FAILS TO DETECT a genome-wide decrease in a histone mark under an inhibitor of its writer: standard methods see no H3K27me3 reduction under an EZH2 inhibitor, and a substantial reduction appears only with a Drosophila chromatin spike-in. Every number in this entry is exactly the kind of normalisation that paper shows to be blind. Egan B et al. 2016
    • The system is murine MLL-AF9 (KMT2A::MLLT3) myeloid cells carrying a retrovirally over-expressed FLAG-tagged construct. This graph exists for human infant B-lineage KMT2A::AFF1, and this corpus's commonest documented synthesis failure is generalising a murine MLL-AF9 myeloid result to that disease. Neville D et al. 2026
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    SGC0946's flat profile is not flatness from inaction: the DMSO-decile-matched null from outside S_FLAG gives 0.2489 / 0.2493 / 0.2483 / 0.2586, so the fusion's own sites lose roughly ten percentage points more H3K79me2 share than matched chromatin does, uniformly across all four quartiles (residual -0.109 / -0.104 / -0.103 / -0.123).

    What this evidence cannot show

    The residual's own Spearman is +0.200, i.e. no monotone structure. With 20 draws the smallest attainable empirical p is 0.0476 and the observed SGC span's p is 0.9048, so the null is entirely consistent with the observed SGC span - which is the intended reading. THE MATCHED NULL CANNOT DISTINGUISH a genuinely uniform drug effect from a normalisation artefact that removes signal uniformly.

    Evidence that argues against this (1)
    • A ten-point excess depletion at S_FLAG relative to matched bins is itself a structured effect, so 'SGC0946 acts uniformly' is true only with respect to eviction depth and false with respect to fusion occupancy. Iqbal Lab oncofusion knowledge graph 2026
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    Under SGC0946 the bin-level bootstrap 95% interval for the Q1-minus-Q4 span is [-0.0078, +0.0138] and spans zero, with only 73.1% of 2,000 resamples keeping the observed sign; under VTP-50469 the same interval is [-0.651, -0.545] with 100% sign agreement.

    What this evidence cannot show

    POST HOC. A bin-level bootstrap resamples bins within one library and cannot speak to library-to-library variation, which is unmeasurable here. The SGC interval spanning zero is the intended negative result and is reported as such rather than as a failure.

    Evidence that argues against this (1)
  4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    SGC0946 is a selective DOT1L methyltransferase inhibitor whose mechanism is catalytic-site remodelling of the enzyme, so depletion of H3K79me2 under it is an on-target effect on the mark's writer rather than on the fusion's chromatin binding.

    Catalytic site remodelling of the DOT1L methyltransferase by selective inhibitors. · Yu W, Chory EJ, Wernimont AK, et al. · Nat Commun 2012
    PMID 23250418DOI
    What this evidence cannot show

    Cited for the compound's identity and mechanism only. That paper characterises the inhibitor biochemically and structurally; it does not contain any of the measurements in this entry, and the 5 uM 72 h exposure used in GSE260742 is far above its biochemical potency, so off-target activity at that concentration is not excluded by the citation.

    Evidence that argues against this (1)
    • A selective inhibitor used at 5 uM for 72 hours is not necessarily selective at that exposure, and 'on-target' is an inference from the compound's published profile, not a measurement in these cells. Daigle SR et al. 2013
Connected to