Independent positive control: of 27 samples, the ONE that calls zero oncoprotein-target promoters is 152985 - the AML that arose during menin-inhibitor treatment, which the paper reports as having oncoprotein and menin reduced to undetectable levels
The published paper says one particular patient sample lost its fusion protein from chromatin entirely after drug treatment. Rebuilt from the raw data with a different pipeline and different genomic regions, that same sample is the only one of twenty-seven that scores zero.
The same thing, for a scientist
**Why a positive control was needed.** Everything else in this beat is a statement about a re-implementation: a different interval set, a different promoter universe, a threshold whose calibration sample lives in another accession. Without an anchor, none of it can be distinguished from a pipeline that does not work. **The anchor.** The paper reports that in *"the 152985 sample that lineage switched during treatment with the menin inhibitor, the oncoprotein and menin are reduced to undetectable levels"*, and that *"46 oncoprotein-binding sites were detectable in the B-ALL sample at diagnosis"* (148752) [E2]. **What this pipeline found, with no knowledge of either statement at scoring time:** across all 27 samples, **exactly one calls zero promoters, and it is 152985**. Its pre-treatment partner 148752 calls 345. And the two of them carry the **lowest and second-lowest amplitude of the whole cohort**, A = **0.347** (rank 1 of 27) and **0.389** (rank 2 of 27) [E1]. They are also the only two of the 27 that fail the pre-registered U1 spread check - **which is the correct behaviour for samples whose oncoprotein is gone**, not a defect. **The cohort ordering the paper predicts is reproduced.** 152985 (post-drug AML) below 148752 (pre-drug B-ALL) below everything else. The absolute numbers differ from the paper's - 345 rather than 46 for 148752 - because the intervals differ; **the ordering and the zero do not** [E1]. **A second, independent anchor.** The top promoters ranked by mean score across all 27 samples, with no gene list supplied, are **REEP3** (16.44x), **SENP6** (10.56x), **MBNL1** (9.96x), **RNF220** (9.14x), **CDK6** (8.29x), **ARID1B** (7.43x), **TAPT1** (7.43x), **BAZ2B** (7.16x) - a list that includes several canonical KMT2A-fusion target genes and is not what a noise-dominated statistic returns [E1].
Why this does not rescue the ordering result
A positive control shows the pipeline detects a very large effect - the complete loss of an oncoprotein - where one exists. It says nothing about resolution at the other end, where the question is whether two samples rank thousands of promoters the same way. Those are different demands, and this control satisfies only the first.
It is also a drug result, with its confound named
Read with the corrected sample labels, GSE252378 contains a pre- and post-menin-inhibitor pair from one patient, and the post-treatment aliquot has no detectable oncoprotein on chromatin at promoter resolution. That is on-target pharmacodynamics in a child. It is confounded totally: the two aliquots differ in lineage (B-ALL then AML) as well as in drug exposure, and there is no vehicle arm, so nothing separates the drug from the switch. The observation is recorded as a reproduction of the paper's, not as independent evidence of drug effect.
| script | analyses/q144-unbounded-magnitude/run.py |
|---|---|
| status | run |
| result summary | 152985 (post-menin-inhibitor AML) is the only sample of 27 calling zero promoters; 148752 (pre-treatment B-ALL) calls 345; the pair carries the two lowest amplitudes. |
top promoters by mean score — the numbers behind the chart
[
{
"symbol": "REEP3",
"mean_S": 4.0395525250272115,
"fold": 16.444719842166208
},
{
"symbol": "SENP6",
"mean_S": 3.400206974894221,
"fold": 10.557577810563314
},
{
"symbol": "MBNL1",
"mean_S": 3.316430099129077,
"fold": 9.961963320824
},
{
"symbol": "RNF220",
"mean_S": 3.1928617433362683,
"fold": 9.14423032945938
},
{
"symbol": "CDK6",
"mean_S": 3.05138190812583,
"fold": 8.290056349657979
},
{
"symbol": "ARID1B",
"mean_S": 2.893857089218105,
"fold": 7.4325491025633506
},
{
"symbol": "TAPT1",
"mean_S": 2.89263111425197,
"fold": 7.42623574574202
},
{
"symbol": "BAZ2B",
"mean_S": 2.8392931049777985,
"fold": 7.156693055321087
},
{
"symbol": "SOCS2",
"mean_S": 2.738506635950051,
"fold": 6.673791596420244
},
{
"symbol": "PARP8",
"mean_S": 2.5083572907282896,
"fold": 5.689718555475333
},
{
"symbol": "ZEB2",
"mean_S": 2.4992881077575593,
"fold": 5.654063585398465
},
{
"symbol": "RNASEH2B",
"mean_S": 2.4699900895953113,
"fold": 5.5403998132618755
},
{
"symbol": "CLECL1",
"mean_S": 2.465338815276815,
"fold": 5.522566229652044
},
{
"symbol": "SKIDA1",
"mean_S": 2.4015969883601733,
"fold": 5.283877384422129
},
{
"symbol": "EMB",
"mean_S": 2.337808163034544,
"fold": 5.055340134211035
}
]- A sample with no detectable oncoprotein on chromatin still relapsed as AML, so clearing the oncoprotein from chromatin is not by itself sufficient for disease control - which is the central worry for menin inhibitors and is visible here in one patient.
- Does 152985 call zero on the authors' own intervals too, or is the zero specific to promoters?
- The top-ranked promoter by mean score across the cohort is REEP3, which this corpus mentions in only three entries. Is it a real KMT2A-fusion target or an artefact of the interval?
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
Re-scoring GSE252378's fragment BEDs at TSS +/- 1 kb promoters, exactly one of 27 samples calls zero oncoprotein-target promoters and it is 152985; its pre-treatment partner 148752 calls 345; and the two carry the lowest (0.347) and second-lowest (0.389) score amplitude of the cohort.
GSE252378 - KMT2A oncoproteins induce epigenetic resistance to targeted therapies (171 GSMs, AutoCUT&RUN / AutoCUT&Tag, hg38 fragment BEDs and bigWigs) · Janssens DH, Henikoff S (Fred Hutchinson Cancer Center / Van Andel Institute) · NCBI Gene Expression Omnibus 2024What this evidence cannot show
These two samples are also the two most deeply sequenced in the cohort, so their low amplitude is partly the depth effect this beat measured separately (Spearman of amplitude against log depth = -0.474, p = 0.012). The ZERO call count is not a depth effect - depth raises binomial power and would raise the count - but the amplitude rank of 1 and 2 should not be read as purely biological.
Evidence that argues against this (1)
- The zero is a property of a threshold: 152985 is not devoid of N-biased promoters, it is devoid of promoters clearing N/C > 1.75 at p < 1e-5. A lower threshold would return a non-zero number for it, and the paper's 'undetectable' carries the same qualification. Neither statement is an absolute-occupancy measurement, and no spike-in exists to make one. Janssens DH et al. 2024
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The paper reports that in the 152985 sample, which lineage switched during menin-inhibitor treatment, the oncoprotein and menin are reduced to undetectable levels, while 46 oncoprotein-binding sites were detectable in the paired B-ALL sample at diagnosis.
MLL oncoprotein levels influence leukemia lineage identities · Janssens DH, Duran M, Otto DJ, Wu W, Xu Y, Kirkey D, Mullighan CG, Yi JS, Meshinchi S, Sarthy JF, Ahmad K, Henikoff S · Nature Communications 15:9341 2024What this evidence cannot show
Quoted from the article text served by PubMed Central (PMC11522475). The 46 refers to the authors' own called intervals and is not comparable in level to a promoter count.
Evidence that argues against this (1)
- Because this beat's pipeline was built while the paper's text was available, the control is not blind. The specific numbers were read after scoring, but the direction was foreseeable, and a genuinely blind reproduction would require a deposit whose result this beat did not already know. Janssens DH et al. 2024
the positive control for the call-count statistic
the corrected pre-/post-treatment labelling this result depends on
the oncoprotein is undetectable on chromatin in the post-menin-inhibitor aliquot
- related to →KMT2A::MLLT1 (MLL-ENL), t(11;19)(q23;p13.3)
the lineage-switching patient carries KMT2A::ENL
menin also reduced to undetectable in the same aliquot, per the paper
the fragment BEDs scored
the menin-inhibitor positive control
the untreated switch is the counter-case that makes the zero a drug effect rather than a lineage effect