Oncofusion

start / graph / ana-cycle-83s-threat-is-refuted-on-the-mll1-channel-in-a-full-two-by-two-at-ten-times-the-dot1l-dose

Analysismoderate confidence 3 of 5: One strong study, or several smaller consistent ones.

A full 2x2 on the MLL1 channel in one run - DMSO, EPZ5676 1 uM, VTP-50469 100 nM and both - closes cycle 83's thirty-cycle-old threat: at the fusion's strongest MLL1 sites the menin inhibitor removes 58.3% against the DOT1L inhibitor's 17.5%, a 3.33-fold separation at TEN TIMES the DOT1L dose Q136 could test, with sequencing depth running opposite to the effect in every arm

also called cycle 83 refuted on the MLL1 channel · the MLL1 two-by-two · 58.3 against 17.5 per cent · depth runs opposite to the effect

A long-standing worry about this project's central idea was that a drug aimed at a different target might disturb the same machinery just as much, which would mean the intended drug is not doing anything special. Nobody had ever run both drugs side by side on the relevant measurement. A public dataset turns out to contain exactly that experiment. The intended drug removes more than three times as much, and a technical artefact that could have faked the result runs the wrong way in every arm, so it cannot be the explanation.

The same thing, for a scientist

**Statement.** Cycle 83's standing threat — that a DOT1L inhibitor moves the **MLL1 channel** as much as a menin inhibitor does — has been open for roughly thirty cycles. Q136 could not test it: GSE149183 has no MLL1 ChIP, so it tested the MEN1 channel instead, at EPZ5676 **100 nM**. **GSE173599 contains the missing experiment as a full 2x2**, one lab, one run, one cell line: | arm | fragments | peaks | all DMSO sites kept | **top-10% sites kept** | |---|---:|---:|---:|---:| | DMSO | **10.54 M** | **9,505** | 100% | **100%** | | EPZ5676 1 µM | 13.59 M | 5,874 | 46.5% | **82.5%** | | VTP-50469 100 nM | **14.74 M** | 2,019 | 17.9% | **41.7%** | | both | 14.40 M | 815 | 7.6% | **15.4%** | **`C2` fires. At the fusion's strongest MLL1 sites the menin inhibitor removes 58.3% and the DOT1L inhibitor 17.5% — a 3.33-fold separation — at 1 µM EPZ5676, ten times the dose Q136 could reach, on the channel Q136 could not reach.** Both halves of the threat close in one experiment. **`C3` was checked first and is resolved in the conservative direction.** Q136 found MACS peak count tracking sequencing depth at ρ = −0.80 and fragment size at +0.70. Here **DMSO is the shallowest library (10.54 M) and calls the most peaks (9,505)**, while every treated arm is **deeper** and calls **fewer** — VTP calls 4.7× fewer peaks on 1.4× more fragments. Fragment size runs the same way (DMSO 210 bp, treated 242–257 bp). **Both confounds predict the opposite of what is observed.** **H3K27ac, the same 2x2, barely moves** — 95.9% / 93.7% / 87.4% of its top sites retained, on a library that is also shallowest at DMSO. **This is not a generic chromatin collapse.** **And the combination is not redundant**: 15.4% retained against 41.7% for the menin inhibitor alone, consistent with the deposit's own title — *chromatin changes from degrading the fusion are phenocopied by combined DOT1L and MENIN inhibition*.

Go deeper
What this does and does not settle

Settled: the specific claim that the two drug classes move the MLL1 channel comparably is false at these exposures. That claim has been the single largest standing threat to the menin-interface hypothesis in this corpus, and it is now contradicted by the only head-to-head measurement of it that exists.

Not settled: whether the separation holds at the 5 µM / 72 h exposures the original MLL1-channel work used. This is 1 µM at 4 days. Cycle 116 established that no public deposit carries a menin or MLL1 readout at ≥5 µM, so that version remains untestable — but the gap has narrowed from 50-fold to 5-fold.

Not settled either: whether any of this transfers to KMT2A::AFF1. MOLM13 is KMT2A::MLLT3. A sibling beat this same cycle found that between these two lines, per-gene ordering transfers (Spearman 0.75–0.79) while magnitude does not — and every number here is a magnitude.

Why a peak-retention metric, and what it costs

Peak counts are notoriously sensitive to depth and to MACS's fragment estimate, which is why retention is scored against the DMSO peak set rather than by comparing counts. That choice makes the metric robust to a uniform sensitivity change but leaves it threshold-bound: a site that falls just under significance is scored as lost, identically to a site that lost all occupancy.

The top-10% stratification is what makes it interpretable. Across all DMSO peaks the separation is 46.5% against 17.9% — the same direction but a smaller ratio — because weak peaks are where threshold effects dominate. At the fusion's strongest sites, where a peak call is least marginal, the separation is widest. That ordering is the opposite of what a pure threshold artefact would produce.

Facts and figures
questiondoes a DOT1L inhibitor move the MLL1 channel as much as a menin inhibitor?
depositGSE173599
publicationPMID 35245435, Molecular Cell 2022
cell lineMOLM13 (KMT2A::MLLT3) - NOT the index fusion
designfull 2x2: DMSO / EPZ5676 1 uM / VTP-50469 100 nM / both, day 4, one run
separation at top sitesmenin inhibitor removes 58.3%, DOT1L inhibitor 17.5%, ratio 3.33
C3 depth checkDMSO is the SHALLOWEST library and calls the MOST peaks; every treated arm is deeper and calls fewer. Depth and fragment size both predict the OPPOSITE of the observed effect, so neither can explain it.
what remains untestablethe 5 uM / 72 h version - cycle 116 established no public deposit carries a menin or MLL1 readout at that exposure; the dose gap narrows from 50-fold to 5-fold
absence classmeasured-not-curated
bite backThis entry removes the largest standing threat to the menin-interface hypothesis and thereby removes a reason for scepticism that has usefully disciplined thirty cycles of work. If the separation is really a dose artefact - a saturating menin inhibitor against a partial DOT1L inhibitor - then this entry will have retired a good objection on a bad comparison, and the project will have lost its best critic.
mll1 top decile retained PCT
both15.4both: 15.4VTP_100nM41.7VTP_100nM: 41.7EPZ_1uM82.5EPZ_1uM: 82.5DMSO100.0DMSO: 100
mll1 all peaks retained PCT
both7.600both: 7.6VTP_100nM17.9VTP_100nM: 17.9EPZ_1uM46.5EPZ_1uM: 46.5DMSO100.0DMSO: 100
peaks called
both815.0both: 815VTP_100nM2,019VTP_100nM: 2019EPZ_1uM5,874EPZ_1uM: 5874DMSO9,505DMSO: 9505
fragments
DMSO10,543,773DMSO: 10543773EPZ_1uM13,590,101EPZ_1uM: 13590101both14,401,535both: 14401535VTP_100nM14,737,278VTP_100nM: 14737278
fragment size bp
DMSO210.0DMSO: 210both242.0both: 242EPZ_1uM252.0EPZ_1uM: 252VTP_100nM257.0VTP_100nM: 257
h3k27ac top decile retained PCT
both87.4both: 87.4VTP_100nM93.7VTP_100nM: 93.7EPZ_1uM95.9EPZ_1uM: 95.9DMSO100.0DMSO: 100
mll1 top decile retained PCT — the numbers behind the chart
{
  "DMSO": 100,
  "EPZ_1uM": 82.5,
  "VTP_100nM": 41.7,
  "both": 15.4
}
mll1 all peaks retained PCT — the numbers behind the chart
{
  "DMSO": 100,
  "EPZ_1uM": 46.5,
  "VTP_100nM": 17.9,
  "both": 7.6
}
peaks called — the numbers behind the chart
{
  "DMSO": 9505,
  "EPZ_1uM": 5874,
  "VTP_100nM": 2019,
  "both": 815
}
fragments — the numbers behind the chart
{
  "DMSO": 10543773,
  "EPZ_1uM": 13590101,
  "VTP_100nM": 14737278,
  "both": 14401535
}
fragment size bp — the numbers behind the chart
{
  "DMSO": 210,
  "EPZ_1uM": 252,
  "VTP_100nM": 257,
  "both": 242
}
h3k27ac top decile retained PCT — the numbers behind the chart
{
  "DMSO": 100,
  "EPZ_1uM": 95.9,
  "VTP_100nM": 93.7,
  "both": 87.4
}
Still open
  1. Normalise the two drugs by their own potency. 100 nM VTP-50469 is likely saturating while 1 uM EPZ5676 may not be; a dose-matched comparison in units of cellular IC50 is the experiment that would separate mechanism from exposure.
  2. Score the degron arm. The same deposit removes the fusion outright (dTAG-VHL 500 nM, 3 h) with MEN1, MLL1, DOT1L, AFF4, CDK9 and Pol II ChIP on either side - the remove-the-fusion reference point this thread has never had, and it is already downloaded.
  3. Does the separation survive in a KMT2A::AFF1 line? A sibling beat found ordering transfers between MOLM13 and RS4;11 but magnitude does not, and every number here is a magnitude.
  4. Re-run cycle 83's original bulk overlap metric on this 2x2. Q136 showed that metric fails its negative control; demonstrating that it also reverses this result would close the original claim rather than merely answering a neighbouring one.
Every source behind this entry (1)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.analysis

    In GSE173599, MACS peak tables for MOLM13 MLL1 ChIP at day 4 give 9,505 peaks under DMSO, 5,874 under EPZ5676 1 uM, 2,019 under VTP-50469 100 nM and 815 under both. Scoring 1 kb-binned retention against the DMSO peak set, 46.5%, 17.9% and 7.6% of all DMSO sites are retained; restricted to the strongest decile of DMSO peaks, 82.5%, 41.7% and 15.4% are retained. Total fragments in treatment are 10,543,773 (DMSO), 13,590,101 (EPZ), 14,737,278 (VTP) and 14,401,535 (both), and MACS fragment-size estimates are 210, 252, 257 and 242 bp. The same 2x2 for H3K27ac retains 95.9%, 93.7% and 87.4% of its strongest decile, with DMSO again the shallowest library at 23,037,839 fragments and the most peaks at 34,796.

    What this evidence cannot show

    n = 1 IN EVERY ARM. One library per condition, no replicates, so there is no measurable noise floor and none is imported - this project's floors span ~20x across arms of a single deposit. Every figure is descriptive and no difference here has an error bar. A PEAK CALL IS A THRESHOLD, not an occupancy measurement: a site dropping just below significance scores identically to one losing all signal, and the all-peaks figures (46.5/17.9/7.6) are more threshold-bound than the top-decile ones. MOLM13 IS KMT2A::MLLT3, NOT THE INDEX FUSION, so every conclusion is partner-imported. The exposures are 1 uM EPZ and 100 nM VTP at day 4, not the 5 uM / 72 h of the work that raised the threat. H3K27ac is not a neutral control - it responds to both drugs on its own account - so its stability is context, not validation. No spike-in exists for these libraries, so nothing here is on an absolute scale.

    Evidence that argues against this (3)
    • The two drugs are not at comparable exposures relative to their own potencies, and the comparison silently assumes they are. VTP-50469 at 100 nM is far above its cellular IC50 while EPZ5676 at 1 uM is closer to the range where DOT1L inhibition is partial. A 3.33-fold separation between a saturating dose of one drug and a sub-saturating dose of another is not the same claim as a difference in mechanism, and this entry cannot distinguish them.
    • The 30-cycle-old threat was never about the same experiment. Cycle 83's alarm came from bulk peak-overlap metrics in a different deposit, and Q136 showed that metric fails its own negative control - two antibodies in untreated cells overlap less than one antibody with and without drug. Refuting a claim on a different dataset with a different metric may simply be answering a different question rather than closing the original one.
    • This is the outcome the project wanted, produced by a coordinator who chose the dataset, the metric, the stratification and the threshold. The top-10% stratification in particular was not pre-registered as the primary readout - the pre-registration named 'retention' without specifying the stratum - and it gives a wider separation than the all-peaks figure. That is defended above on threshold grounds and it remains a choice made after seeing both numbers.
Connected to