The menin, ENL and DOT1L replicates of the drug-exposed relapse sample agree with each other at Spearman 0.11 to 0.25, below the floor this project set in the previous cycle, so no quantitative claim from that arm survives
In the sample taken while the child was on the drug, running the same experiment twice gave almost unrelated answers for menin, ENL and DOT1L. Two runs of the same measurement that disagree that badly cannot settle anything - including the published claim that ENL stayed put.
The same thing, for a scientist
Promoter-level Spearman between the two CUT&Tag replicates of each channel, over the 19,404-gene filter-free set. In **148752** (drug-free): menin **0.691**, ENL **0.620**, DOT1L **0.311**. In **152985** (on revumenib): menin **0.246**, ENL **0.247**, DOT1L **0.110**. The previous cycle of this project judged H3K27ac replicates agreeing at 0.21-0.29 to be too low for that arm to carry a conclusion; all three 152985 CUT&Tag channels are at or below that floor, and DOT1L is at less than half of it. The CUT&RUN channels of the same sample hold up better - the two N-terminal reagents agree at **0.568** and the two C-terminal at **0.659** - but both are far below their 148752 values of 0.909 and 0.935. **The drug arm of this pair is instrumentally weak in every channel, and worst in the three channels the published claim rests on.**
The table
| channel | 148752 (drug-free) | 152985 (on revumenib) | what the pair is |
|---|---|---|---|
| menin | 0.691 | 0.246 | two CUT&Tag replicates |
| ENL | 0.620 | 0.247 | two CUT&Tag replicates |
| DOT1L | 0.311 | 0.110 | two CUT&Tag replicates |
| KMT2A-N | 0.909 | 0.568 | two different antibodies |
| KMT2A-C | 0.935 | 0.659 | two different antibodies |
Note that the N and C rows are a harder test than the replicate rows - two different antibodies raised against the same fragment - and they still score higher in both samples. DOT1L's 0.311 in the drug-free sample is already the weakest channel in this deposit and should be treated as such wherever DOT1L is quoted.
The interpretation this cannot settle, stated as a fork
Low replicate agreement has two sufficient causes and this deposit distinguishes neither:
- The libraries are poor. Depth differs sharply within pairs - 152985 ENL R1 has 1.43 million placed fragments and R2 has 5.35 million; DOT1L R1 1.39 million and R2 5.11 million - which alone depresses a rank correlation.
- The target is gone, so both replicates are background. Replicates of background do not correlate, and a collapse to background is precisely what the paper reports for this sample.
Reading 2 is a result and reading 1 is a defect, and they produce the same number. Every claim this beat makes about 152985 is written to survive both readings: it says what is not above background, never how much of something is present.
What this does to the published ENL claim
The paper states that in the menin-inhibitor-treated AML "ENL remained bound at appreciable levels over numerous canonical oncoprotein-target genes". That claim rests on the two 152985 ENL CUT&Tag libraries, which agree with each other at 0.247. It survives here only in a direction-of-change form, which is robust to this defect because both replicates give the same sign - see ana-q143-enl-redistributes-to-the-gmp-like-peaks-rather-than-simply-persisting.
| prior project floor | 0.21-0.29 (Q140, SEM H3K27ac, GSE159608) |
|---|---|
| the fork | low agreement is produced both by a poor library and by a genuinely absent target; this deposit cannot separate them |
| consequence | no quantitative 152985 claim is made in this beat; only statements of the form 'not above background' |
| status | run |
| script | analyses/menin-under-revumenib-in-an-infant-patient-pair/run2_controls.py |
replicate spearman
{
"148752": {
"MENIN": 0.691,
"ENL": 0.62,
"DOT1L": 0.311,
"KMT2A_N_two_antibodies": 0.909,
"KMT2A_C_two_antibodies": 0.935
},
"152985": {
"MENIN": 0.246,
"ENL": 0.247,
"DOT1L": 0.11,
"KMT2A_N_two_antibodies": 0.568,
"KMT2A_C_two_antibodies": 0.659
}
}placed fragments within pair
{
"152985_ENL": [
1426687,
5348513
],
"152985_DOT1L": [
1389218,
5109006
],
"152985_MENIN": [
5526235,
4943825
]
}- A published claim built on libraries that disagree with each other at 0.247 will be cited for years; the correct response is to ask for the replicate correlation, not to assume the conclusion is wrong.
- If a clinical pharmacodynamic assay is designed from this pair, note that its drug-arm replicates did not reproduce - an assay that cannot reproduce itself in the treated state is not a pharmacodynamic assay.
- Does down-sampling the deeper replicate of each 152985 pair to the shallower one's depth change the agreement? That would separate the depth explanation from the biology explanation and was not run here.
- Is DOT1L's 0.311 in the drug-free sample a property of the D4O2T antibody, of DOT1L's broad gene-body distribution, or of the library? DOT1L is the weakest channel in both samples.
- Should this project adopt a stated replicate-agreement floor as a field in every reanalysis entry, rather than as a convention recalled from the previous cycle?
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
The three CUT&Tag channels of the revumenib-exposed sample have promoter-level replicate Spearman of 0.246 (menin), 0.247 (ENL) and 0.110 (DOT1L), against 0.691, 0.620 and 0.311 in the drug-free sample of the same patient.
analyses/menin-under-revumenib-in-an-infant-patient-pair - pre-registered re-analysis of GSE252378 samples 148752 and 152985 · Q143-menin-under-drug-in-infants (Oncofusion Knowledge Graph) · this repository 2026What this evidence cannot show
Spearman over a fixed filter-free promoter set; the pairs differ in depth by up to 3.8-fold within a sample, which depresses a rank correlation independently of biology.
Evidence that argues against this (1)
- A collapse in replicate agreement is the expected consequence of the target genuinely leaving chromatin, so treating it purely as a quality defect would discard the paper's finding by assuming it away. Janssens DH et al. 2024
- E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
The previous cycle of this project treated replicate agreement of 0.21-0.29 as too low for an arm to carry a conclusion.
analyses/menin-under-revumenib-in-an-infant-patient-pair - pre-registered re-analysis of GSE252378 samples 148752 and 152985 · Q143-menin-under-drug-in-infants (Oncofusion Knowledge Graph) · this repository 2026What this evidence cannot show
That threshold was set on a different deposit (GSE159608, SEM H3K27ac) and is a convention adopted by this project, not a field standard. It is used here as a pre-existing line rather than one drawn after seeing these numbers.
Evidence that argues against this (1)
- A threshold adopted by this project on one deposit is a convention, not a field standard, and CUT&Tag replicate correlations in primary patient material are routinely lower than in cell lines - so applying a cell-line-derived floor to a patient sample may condemn libraries that are normal for their material. Kaya-Okur HS et al. 2019
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset
The library depths within the 152985 CUT&Tag pairs differ by up to 3.8-fold (ENL 1.43M vs 5.35M placed fragments; DOT1L 1.39M vs 5.11M).
analyses/menin-under-revumenib-in-an-infant-patient-pair - pre-registered re-analysis of GSE252378 samples 148752 and 152985 · Q143-menin-under-drug-in-infants (Oncofusion Knowledge Graph) · this repository 2026What this evidence cannot show
Counted directly from the deposited hg38 BED files; 'placed' means a fragment midpoint landing on a primary chromosome.
Evidence that argues against this (1)
- A depth difference within a pair depresses a rank correlation but does not by itself make a library uninformative, and the deeper member of each pair may be perfectly good; the correct response is to down-sample and re-measure, which was not done here. Janssens DH et al. 2018
the agreement is measured within the pair
the two instrument checks that gate everything downstream
the deposit's two structural weaknesses in the drug arm
the drug arm's replicate agreement bounds what can be claimed
the drug arm's replicate agreement bounds this result
the drug arm's replicate agreement bounds the magnitude