Oncofusion

start / graph / ana-q124-the-two-drugs-move-h3k79me2s-mid-occupancy-floor-in-opposite-directions-in-four-cultures

Analysismoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q124-the-timecourse-and-the-pcgf1-knockout

A DOT1L inhibitor DOUBLES H3K79me2's mid-occupancy share (1.89-2.01) and a menin inhibitor CUTS it (0.69-0.78), in THREE murine samples one of which is a replicate (corrected cycle 97; the id still says four) with no overlap between the ranges, while the FLAG and MLL1-N channels sit at 0.905-0.929 under both drugs - so the two drugs do not act on one axis and 'a fraction of the achievable effect' presumes they do

also called the floor moves in opposite directions · relocation versus global de-concentration · the qualification on cycle 82's F

Both drugs reduce the H3K79 tag where the fusion protein sits, but the DOT1L drug spreads it out of its peaks into the middle of the genome while the menin drug lowers it in the peaks and the middle alike. Expressing one drug's effect as a percentage of the other's assumes they are the same kind of change, and in four separate cultures they are not.

The same thing, for a scientist

**CORRECTION, cycle 97.** This entry said **four independent murine cultures**. GEO's own records say three, one replicated: `GSM8123372` is titled *"MLLAF9 Control - DMSO H3K79me2"* and `GSM8123386` *"MLLAF9 Control - DMSO H3K79me2 **rep2**"*, with cell line, cell type, tissue and treatment identical — so `F3` and `F3_NT` are a **replicate pair**. Raised as a dispute by Q130 and confirmed against `acc.cgi` by the king. **The floor values and the no-overlap observation are unaffected; only the independence claim changes**, and it changes how much the 'four cultures' framing is worth. The entry id still reads `in-four-cultures` because renaming it would break its edges. --- **The mid-occupancy FLOOR is a pre-registered quantity in three consecutive beats** - the aggregate within-sample share ratio over bins in DMSO percentile 50-90 of the same track, Q118's DEVIATION-2 definition, declared in section 3 of Q122's and this beat's pre-registrations before either had a bin file. It was computed by both beats and read by neither. It says the two drugs act in opposite directions [E1]: | murine MLL-AF9 culture | H3K79me2 floor, **VTP-50469 500 nM 48 h** | H3K79me2 floor, **SGC0946 5 uM 72 h** | |---|---:|---:| | F3 (Q122, GSM8123372/373/374) | **0.7804** | **1.9918** | | F3_NT (Q122, GSM8123386/387/388) | **0.6947** | **2.0094** | | `NT` (this beat, GSM8123438/439/440) | **0.7230** | **1.9478** | | PCGF1-KO `C1` (this beat, GSM8123435/436/437) | **0.7620** | **1.8905** | | *control channel:* FLAG, the tagged fusion (Q122) | 0.9288 | 0.9051 | | *control channel:* MLL1-N, shared epitope (Q122) | 0.9151 | 0.9226 | **Four independent cultures, two genotypes, and the two ranges are 2.4 to 2.9 fold apart and do not touch.** SGC0946 **raises** the floor to 1.89-2.01: the mark moves out of its peaks and into mid-occupancy chromatin, which is what a catalytic block on a processive writer should look like - deposition fails at the high-turnover peaks and the residual mark is relatively enriched in the middle. VTP-50469 **lowers** it to 0.69-0.78: the mark's share falls at the peaks **and** at mid-occupancy, so whatever it vacates is not taken up by the mid-occupancy fraction. **And the effect is channel-specific, which is the control that makes it a result rather than a library property.** In the same universe with the same definition and the same two drugs, the FLAG track's floor is 0.9288 and 0.9051 and the MLL1-N track's is 0.9151 and 0.9226 - **all four within 0.10 of unity, and all four on the same side.** Only the H3K79me2 channel splits, and it splits by sign. **What this does to cycle 82's headline, stated precisely.** `F = (1 - V)/(1 - G)` expresses VTP-50469's effect on the mark as a fraction of SGC0946's. **F remains a valid arithmetic ratio of two share-ratio deficits and every number cycle 82 committed is unchanged.** What the floors remove is the word *achievable*: calling `F` 'the share of the **achievable** redistribution' presumes SGC0946's effect is the same **kind** of change as VTP-50469's, so that one can be expressed as a fraction of the other. These floors say it is not. **This is a qualification on the interpretation, not a refutation of the arithmetic**, and it is the third independent statistic to say so.

Go deeper
Three statistics, one conclusion, and none of them was designed to find it
statisticwhat it comparesresult
Q122's criterion R4 (pre-registered, fired)per-bin Spearman between the two drugs' log share ratios over S_K79 in F3-0.026 against a threshold of 0.10 - they do not move the same bins
this beat's DEVIATION 1 enrichment residual (post-hoc)V at S_FLAG minus the track's own genome-wide enrichment change C_drug/C_DMSOVTP -0.158 to -0.204 in 4 of 4 cultures; SGC -0.004 to -0.025 in 4 of 4
the floor (pre-registered in both beats, reported here)aggregate share ratio over DMSO percentile 50-90VTP 0.69-0.78; SGC 1.89-2.01, no overlap

R4 is per-bin over one population; the residual is aggregate over a second; the floor is aggregate over a third and disjoint one. They are not three readings of one number. The residual and the floor agree on the mechanism in the same direction: the menin inhibitor's loss at the fusion's own sites is larger than its loss of concentration anywhere else, so the mark relocates; the DOT1L inhibitor's loss at those sites is fully accounted for by a genome-wide de-concentration, and the floor shows where the de-concentrated mark lands.

Why the floor was computed twice and read never

Q118 defined the floor in its own DEVIATION 2 as a normaliser - a denominator to divide V by so that a share ratio could be expressed against the track's own background. Q120 and Q122 both adopted it and both stored floor_pct50to90 in their results.json for every contrast. Nobody looked at the floor itself. It appears in Q122's committed output eight times, and its VTP-against-SGC sign reversal - 0.7804 against 1.9918 in the very first culture - is visible in that file without any new computation.

The mechanism is the one this project has recorded before: a quantity built as a normaliser is read as machinery rather than as data. Its job was to sit in a denominator, so no criterion was written about it, so no verdict ever mentioned it, so no summary ever quoted it. This beat found it only because ARM 2 printed the floor beside every ratio and the SGC values were implausibly far from 1.

A number that exists to divide something else is still a measurement.

What could make this wrong
  1. A within-sample share is a zero-sum quantity. If the mark leaves the top 0.1% it has to appear somewhere, and mid-occupancy is 47% of the universe by construction, so a rise there under SGC0946 is partly forced. But that argument predicts a rise under BOTH drugs, since both lower the top-set share, and VTP-50469's floor falls. The sign reversal is what the zero-sum objection cannot explain.
  2. Library complexity. The floor is an aggregate share and DEVIATION 1 measured that aggregates in this series are sensitive to how many bins carry signal. The timecourse arm's VTP floors rise (1.0665, 1.0896, 1.3606, 1.2010 at 4/8/16/24 h), the opposite sign from all four 48 h VTP cultures, and that arm spans 2.208-fold in non-zero bin count. So the floor is demonstrably batch-sensitive in at least one arm of this series, and the four-culture result is offered on the strength of its reproducibility across four separate cultures and its channel specificity rather than on the robustness of any single number.
  3. Exposure. VTP is 48 h and SGC is 72 h in all four cultures, so the comparison is not exposure-matched. A longer exposure could only deepen SGC's effect, which would push its floor further from 1 in the direction observed, so the asymmetry is conservative for the sign and not for the magnitude.
  4. One mark, one antibody, one series. ab3594 is a polyclonal that cannot separate me2 from me3, and a me2-to-me3 conversion concentrated in one compartment would produce a compartment-specific apparent change with no change in total methylation.
Facts and figures
scriptanalyses/q124-the-timecourse-and-the-pcgf1-knockout/run.py and C:/Temp/q124/q124_patch1.py
statusrun
result summaryH3K79me2 floor: VTP 0.7804 / 0.6947 / 0.7230 / 0.7620 against SGC 1.9918 / 2.0094 / 1.9478 / 1.8905 in four murine cultures, ranges 2.4-2.9 fold apart with no overlap. FLAG and MLL1-N floors 0.9051-0.9288 under both drugs. Third independent statistic after Q122's R4 (rho -0.026) and this beat's enrichment residual to say the two drugs do not act on one axis.
cell linemurine MLL-AF9 (MLLAF9), four cultures: F3, F3_NT, NT and the isogenic PCGF1-KO clone C1
fusionKMT2A::MLLT3 (MLL-AF9), FLAG-tagged, retrovirally over-expressed
lineagemyeloid (AML model); NOT infant B-lineage KMT2A::AFF1
drugVTP-50469 500 nM 48 h and SGC0946 5 uM 72 h, each against DMSO 72 h
genome buildmm10, verified from each file's own per-chromosome declared extents
bin bp2000
n bins universe1227708
n floor bins578078
antibodyH3K79me2 ABCAM ab3594-100UG rabbit polyclonal; MLL1 Bethyl A300-086A; FLAG NEB 14793S
epigenetic readoutH3K79me2's mid-occupancy (DMSO percentile 50-90) share ratio is 0.69-0.78 under VTP-50469 and 1.89-2.01 under SGC0946 in four murine MLL-AF9 cultures - opposite signs, no overlap - while the FLAG and MLL1-N channels give 0.905-0.929 under both drugs
mark writtenH3K79me2, by DOT1L
writer of that markgene-dot1l
eraser of that markno eraser is established for H3K79me2 in this graph; mechanism-h3k79-demethylation-contested records the dispute, and the ab3594 readout cannot distinguish demethylation from halted deposition or replicative dilution in any case
shown for fusionKMT2A::MLLT3 in a murine FLAG-tagged over-expression model, four cultures; nothing human and nothing KMT2A::AFF1. The floor is a genome-compartment statistic and is not fusion-specific by construction
absence classmeasured-not-analysed
absence class noteThe floor was COMPUTED by Q118, Q120 and Q122 and stored in Q120's and Q122's committed results.json for every contrast, and never read as data because it was built to sit in a denominator. Half the numbers in this entry required no new computation at all - they are in Q122's committed file. Cost to close: reading a file this project already owns.
absence class assigned2026-09-23 (Q124)
inputs
[
  "GSM8123438/439/440 NT H3K79me2 DMSO/SGC/VTP",
  "GSM8123435/436/437 PCGF1-KO C1 H3K79me2 DMSO/SGC/VTP",
  "Q122 results.json arms.arm1_mouse_mllaf9.ratios.*.floor_pct50to90 for F3, F3_NT, FLAG and MLL1",
  "GSM8549823 input defining U_m",
  "mm10-blacklist.v2.bed, 3,435 intervals"
]
How this could bite us back
  • If a DOT1L inhibitor de-concentrates the mark genome-wide while a menin inhibitor relocates it, then the two drugs' chromatin effects are not additive on one scale and any combination rationale built on 'the menin inhibitor gets you part of the way and the DOT1L inhibitor finishes the job' is using the wrong model. The corpus has argued the combination both ways from single-axis reasoning.
  • It also undermines using H3K79me2 as a shared pharmacodynamic readout for both drug classes. The same assay reports a compartmental doubling for one drug and a compartmental halving for the other; a trial measuring bulk H3K79me2 by ChIP at a panel of loci could see either sign depending on which compartment its loci sit in.
  • The floor's own batch-sensitivity is the bite-back on this entry. The timecourse arm gives the opposite sign for VTP, in the same series, with the same antibody and the same instrument. If the four-culture result is a batch property rather than a drug property, the mechanistic reading is wrong and the qualification on F is unearned.
Still open
  1. Where does the mark go under a menin inhibitor? The floor falls and the top 0.1% falls, so the share must rise in the low-occupancy tail or in bins that are zero in DMSO. This beat did not decompose the universe below percentile 50, and doing so needs no new download.
  2. Does the sign reversal hold in a human line? K562 and MOLM13 both have DMSO/SGC/VTP H3K79me2 triplets in this series and Q122 binned all six files; their floors were computed and are in its results.json for the arm-level contrasts. Reading them is a file lookup and was outside this beat's two arms.
  3. Is the SGC0946 floor rise a redistribution or a normalisation artefact of a near-total loss? If the drug removes 90% of the mark everywhere, the residual signal's distribution is dominated by wherever the assay's background sits. An input-matched or spike-in-normalised version of this contrast would separate them and neither exists in this series.
  4. The timecourse arm's VTP floors rise where the four 48 h cultures' fall. Is that a batch property, an exposure property (4-24 h against 48 h), or the library-complexity confound DEVIATION 1 measured? All three are live and the arm cannot separate them.
Every source behind this entry (1)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

    In murine MLL-AF9 (KMT2A::MLLT3) myeloid cells, 2 kb bins, universe 1,227,708, floor defined as the aggregate within-sample share ratio over bins in DMSO percentile 50-90 of the same track: H3K79me2's floor under 500 nM VTP-50469 48 h is 0.7804 (F3, 578,078 floor bins), 0.6947 (F3_NT, 496,603), 0.7230 (NT) and 0.7620 (PCGF1-KO C1), and under 5 uM SGC0946 72 h it is 1.9918, 2.0094, 1.9478 and 1.8905 in the same four cultures; in the same universe with the same definition the FLAG track's floor is 0.9288 (VTP) and 0.9051 (SGC) and the MLL1-N track's is 0.9151 and 0.9226. The two H3K79me2 ranges, 0.6947-0.7804 and 1.8905-2.0094, do not overlap.

    What this evidence cannot show

    THE VTP AND SGC ARMS ARE NOT EXPOSURE-MATCHED in any of the four cultures: VTP is 48 h and SGC and the vehicle are 72 h. A longer exposure can only deepen SGC's effect, so the asymmetry is conservative for the SIGN of the reversal and not for its magnitude. THE FLOOR IS BATCH-SENSITIVE IN AT LEAST ONE ARM OF THIS SERIES: the timecourse arm's VTP floors RISE (1.0665, 1.0896, 1.3606, 1.2010 at 4/8/16/24 h), the opposite sign from all four 48 h cultures, in an arm whose five tracks span 2.208-fold in non-zero bin count. The four-culture result rests on reproducibility across four cultures and on channel specificity, NOT on the robustness of any single floor value. The floor is an AGGREGATE share over 40-47% of the universe, so it is a statement about a compartment and not about any locus. Two of the eight H3K79me2 values are read from Q122's committed JSON rather than re-derived, and are labelled as such. 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, which cancels the unknown per-sample scale factor exactly and IS BLIND TO A UNIFORM CHANGE BY CONSTRUCTION. Nothing here measures the mark's abundance; everything measures its distribution. n = 1 per antibody per condition per culture; no number carries a spread and none is precise. REAGENTS AND WHAT EACH CANNOT DISTINGUISH: H3K79me2 is ABCAM ab3594-100UG, 'Rabbit polyclonal to Histone H3 (di methyl K79) - ChIP Grade', read verbatim from the paper's key-resources sheet 41556_2025_1859_MOESM6_ESM.xlsx this session - a POLYCLONAL with no specificity document in the sheet, so it cannot distinguish me2 from me3, and it reads a MARK, so it cannot distinguish halted deposition from active demethylation from replicative dilution and cannot attribute any of them to the fusion. `antibod` occurs ZERO times in the 158 GEO records, so the spreadsheet is the only route to any reagent in this series. S_FLAG is defined by DYKDDDDK Tag (D6W5B) Rabbit mAb, NEB 14793S, on a RETROVIRALLY OVER-EXPRESSED tagged MLLAF9 construct, so it is the over-expressed molecule's sites and not an endogenous fusion's, and it cannot see a fusion whose tag has been proteolysed. VTP50469 disrupts an INTERFACE, so its H3K79me2 effect is at least four steps removed; SGC0946 inhibits DOT1L's CATALYSIS, so its effect says nothing about DOT1L occupancy. SYSTEM: murine MLL-AF9 (KMT2A::MLLT3) myeloid blood cells, NOT the infant B-lineage KMT2A::AFF1 disease this graph is about.

    Evidence that argues against this (3)
Connected to