Both datasets this beat used carry an exposure mismatch that could have produced its results, and both were measured rather than argued away: MOLM13's MENIN pair is 4 days against everything else's 3, and GSE260742's VTP arms are 48 h against 72 h DMSO controls - the first moves a criterion by 0.024 and the second is matched between the two arms being compared
also called the exposure hazards · d3 versus d4 · 48h versus 72h
In both datasets some samples were treated for longer than others, which could easily have created the differences we found. Checking each one shows neither explains the result.
The same thing, for a scientist
**HAZARD 1 - GSE127507 MOLM13: the MENIN pair is a 4-day exposure and the other eight samples are 3 days** [E1]. `timepoint: 4 days` on GSM3635683 and GSM3635684; `timepoint: 3 days` on the DOT1L, MLL1n, H3K79me2 and input pairs. A longer exposure can only push menin further down, so a *low* menin figure is uninterpretable as biology and a *high* one would be conservative. Handled by pre-registering two region sets and scoring every criterion on both: | | MENIN | DOT1L | H3K79me2 | MLL1-N | |---|---:|---:|---:|---:| | **S_AB**, MLL1-N and MENIN tops (mixes d3 and d4 in its own definition), n = 467 | 0.1657 | 0.2219 | 0.8667 | 0.8020 | | **S_d3**, MLL1-N and DOT1L tops (pure 3-day), n = 565 | 0.1866 | 0.2669 | 0.9546 | 0.8533 | **Removing the 4-day track from the set definition moves the menin/DOT1L difference from 0.056 to 0.080** - across a band edge, which is why criterion O7 reads DOES NOT FIRE - and **changes no ordinal verdict**: the order is `MENIN < DOT1L < MLL1-N < H3K79me2` on both. **HAZARD 2 - GSE260742: the VTP arms are 48 h and their DMSO controls are 72 h** [E2]. `treatment: 48hrs` on GSM8123361 (FLAG VTP) and GSM8123397 (MLL1 VTP); `treatment: 3 days` on GSM8123359 (FLAG DMSO) and `treatment: 72hrs` on GSM8123395 (MLL1 DMSO). So neither arm is internally symmetric - **and the two arms carry the identical asymmetry, so the FLAG-versus-MLL1 comparison, which is the only comparison ARM 2 makes, is exposure-matched to itself.** My dispatch brief did not mention the 48 h/72 h asymmetry at all. **HAZARD 3 - the dose.** GSE260742 is **500 nM** VTP50469; all 26 GSE127507 samples are **330 nM**. These are not matched doses and no claim in this beat describes them as matched. My brief flagged this one correctly.
A fourth mismatch the brief did not name and neither did I, until after the fact
The FLAG files are GSM8123359_ChIP_FLAG_F3_DMSO.filt.bam.tdf and the MLL1 files GSM8123395_ChIP_MLL1_F3_NT_DMSO.tdf. F3 against F3_NT, and .filt.bam.tdf against .tdf.
Both sample titles say 'MLLAF9 Control', so the depositors regard them as the same condition, and Q117 measured both naming families as batch-1 low-artefact files - 1.65% and 1.74-1.89% of coverage in mm10 blacklist intervals respectively, against 14-46% for this series' batch-2 trimmed_* files. So the processing states are comparable in the way that matters here.
But 'the same cells' is a premise and not a measurement. NT most likely denotes a non-targeting-guide control population, and if the FLAG ChIPs were done on the parental culture and the MLL1 ChIPs on an NT-transduced derivative, then the phrase 'an MLL1 antibody and a FLAG tag in the SAME cells' - which is how this beat's own dispatch brief framed the experiment, and how Q117 framed it before that - is weaker than it sounds. This was not pre-registered and is recorded as a deviation. It does not change any number; it changes how firmly the numbers can be attributed to one nucleus.
| status | run |
|---|---|
| result summary | Hazard 1 (MOLM13 MENIN d4 against d3) moves one criterion by 0.024 across a band edge and no ordinal verdict. Hazard 2 (48 h VTP against 72 h DMSO in GSE260742) is identical in both arms being compared. Hazard 3 (500 nM against 330 nM) is real and no claim treats the doses as matched. |
| drug | VTP-50469 330 nM (GSE127507) and VTP50469 500 nM (GSE260742) - NOT matched |
| shown for fusion | KMT2A::MLLT3 and KMT2A::AFF1 in human, MLL-AF9 in mouse - three systems, three exposure schedules, all transcribed from the records |
| absence class | measured-not-reported |
| absence class note | The depositors know whether the `F3` and `F3_NT` ChIPs came from the same culture and no record says so, and they know how many divisions the treated cultures completed and no record says that either. Both are facts the assay generated and nobody wrote down; the deposited accession that would carry them is GSE260742 itself. |
| absence class assigned | 2026-09-23 (Q120) |
- An exposure mismatch inside a single deposited panel is the cheapest possible way to manufacture a between-antibody difference, and it is invisible in every figure. This beat found two in two series and only one of them had been noticed before.
- Are the F3 and F3_NT populations the same culture? One line in a sample record would settle it and 158 records do not contain it.
- Why is MOLM13's MENIN pair a day longer than everything else in its own series? A 4-day sample in a 3-day panel is usually a repeat, and if it is, the original 3-day MOLM13 MENIN data may exist unpublished.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
GSM3635683 and GSM3635684 (MOLM13 MENIN) carry `timepoint: 4 days` and the other eight MOLM13 samples carry `timepoint: 3 days`; all 26 GSE127507 samples carry 330 nM VTP-50469 or DMSO 0.33%. Scoring the MOLM13 ordering on a pure-3-day region set instead of the mixed set moves the menin/DOT1L difference from 0.056 to 0.080 and changes no ordinal verdict.
GSE127507 per-sample SOFT records, 26 samples, 134,259 characters, fetched live 2026-09-23 · NCBI GEO 2019What this evidence cannot show
The pure-3-day set S_d3 still SCORES the 4-day MENIN track; it only removes it from the set DEFINITION. No construction available in this series removes the 4-day exposure from a MOLM13 menin measurement, because there is no 3-day MOLM13 MENIN sample. NO SPIKE-IN NORMALISATION. All 26 GSE127507 records state a Drosophila spike-in was run, measured, and deliberately not used, and no scale factor is deposited; no GSE260742 record contains spike-in vocabulary at all. Every cross-condition number here is therefore a WITHIN-SAMPLE share or rank, which cancels the unknown per-sample scale factor exactly. No raw fold change is reported as a biological quantity anywhere in this beat.
Evidence that argues against this (1)
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
In GSE260742, GSM8123361 (FLAG VTP50469 500 nM) and GSM8123397 (MLL1 VTP50469 500 nM) both carry `treatment: 48hrs`, while GSM8123359 (FLAG DMSO) carries `treatment: 3 days` and GSM8123395 (MLL1 DMSO) carries `treatment: 72hrs`; the FLAG files are named `*_FLAG_F3_*.filt.bam.tdf` and the MLL1 files `*_MLL1_F3_NT_*.tdf`.
GSE260742 per-sample SOFT records, 158 samples, 419,824 characters, fetched live 2026-09-23 · NCBI GEO 2025What this evidence cannot show
Whether `F3` and `F3_NT` are the same culture is not stated in any record; both titles read 'MLLAF9 Control'. This is an inference from filenames and is flagged as such.
Evidence that argues against this (1)
the pre-registered exposure hazard does not produce the ordering result: the pure 3-day set gives the same ordinal answer
the GSE260742 exposure asymmetry is identical in the FLAG and MLL1 arms, so the comparison between them is matched; the 500 nM against 330 nM dose difference is not
the batch-1 classification holds for the MLL1 files: 1.74-1.89% of coverage in blacklist intervals, against 14-46% for the batch-2 trimmed_* files