The oncoprotein-target site count for sample 148752 is 46, 345 or 745 depending on the interval set and the universe, and the corpus has been quoting one of the three with no label
How many genes a cancer protein sits on depends entirely on which stretches of DNA you decided to count. The same sample gives 46, 345 or 745 depending on that choice, so a bare count means nothing without saying which regions were used.
The same thing, for a scientist
The paper states that **46 oncoprotein-binding sites** were detectable in 148752, the KMT2A::ENL B-ALL at diagnosis, before menin-inhibitor treatment. q144's re-implementation of the same 1.75 / 1e-5 criterion on TSS+/-1kb refGene promoters restricted to the 27-way common set reports **345**. This beat's re-implementation on the pair universe for 148752/152985 (17,254 promoters measurable in both, 88.9% of the promoter set) after within-pair depth matching reports **745**. All three are correct for their own intervals. The corpus has carried '345' in a brief as though it were the sample's site count, and the paper's own 46 was not beside it.
The three counts and what each one is
| count | intervals | universe | source |
|---|---|---|---|
| 46 | the paper's own called peak intervals, calibrated against a CD34+ oncoprotein-negative sample | genome-wide | the publication |
| 345 | TSS+/-1kb refGene promoters | the 4,646 promoters covered in all 27 samples | q144 |
| 745 | TSS+/-1kb refGene promoters | the 17,254 promoters covered in both members of this pair | this beat, depth-matched |
The 16-fold spread between the paper's figure and this beat's is not a disagreement. A promoter window is not a peak; a 27-way intersection is not a pair universe; and the paper's calibrator is not in the deposit. A count quoted without its intervals, its universe and its calibrator is not a measurement.
The claim that does survive all three
152985 - the AML the same patient's disease switched into, sampled during revumenib - calls zero promoters under every version. The paper says the N- and C-terminal MLL signal was reduced to undetectable levels; q144 reported 0; this beat reports 0 in all three universes. That is the one number in this family that is robust to every analysis choice tested, and it is the reason the 148752/152985 pair is withheld rather than reported as a Jaccard.
The standing consequence for this corpus
Any entry quoting an oncoprotein-target site count from this deposit must name (a) the interval set, (b) the universe or coverage filter, and (c) whether a fusion-negative calibrator was used. Where the paper reports its own figure for the same sample, quote both and say which is which. This entry exists so that the next beat reaching for '345' finds '46' in the same sentence.
| script | analyses/q146-one-universe-jaccard/run146.py |
|---|---|
| status | run |
| result summary | 148752 site counts: 46 (paper, own peak intervals, CD34+ calibrated), 345 (q144, refGene promoters, 27-way universe), 745 (this beat, refGene promoters, pair universe, depth-matched). 152985 calls 0 under all three. |
| count paper | 46 |
| count q144 | 345 |
| count q146 pair universe | 745 |
| count 152985 | 0 |
| pair universe | 17254 |
| comparable fraction | 0.8892 |
inputs
[ "analyses/q146-one-universe-jaccard/results.json" ]
- A site count is the currency in which this corpus states how much oncoprotein a sample carries and therefore how much menin-inhibitor target there is. A 16-fold interval-dependent spread in that currency means a cross-study comparison of 'how many targets' is meaningless without harmonising intervals first.
- Would applying a CD34+ oncoprotein-negative calibrator - available in GSE159608 but in a different accession and a different protocol - bring the promoter-window count toward the paper's 46, and if so by how much? That is the single most informative check remaining on this criterion.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The paper reports that 46 oncoprotein-binding sites were detectable in the KMT2A::ENL B-ALL sample at diagnosis, and that menin-inhibitor treatment reduced the genome-wide MLL N- and C-terminal signal in the paired AML to undetectable levels.
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
Read from the article full text. The figure is for the paper's own peak intervals with its CD34+ calibration sample; the text does not state the interval width or the exact calibrator normalisation in the sentence quoted.
Evidence that argues against this (1)
- A count of 46 in a sample that this beat's promoter-window implementation calls 745 times could alternatively mean the paper's calibrator removes most promoter-proximal calls as non-specific, in which case 745 is not a looser version of 46 but a different quantity entirely. Janssens DH et al. 2024
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset
The fragment-level data that produce the 345 and 745 counts are the deposited hg38 fragment BEDs of GSE252378 for samples 148752 and 152985.
GSE252378: AutoCUT&RUN of KMT2A-rearranged primary leukemia samples (KMT2A N- and C-terminal antibodies, ENL, DOT1L, H3K4me3, IgG) · Janssens DH, et al. · NCBI Gene Expression Omnibus 2024What this evidence cannot show
148752 is the B-ALL at diagnosis, BEFORE menin-inhibitor treatment; GEO's 'treated with the Menin inhibitor' string is a case-level label attached to both libraries and must not be read as an exposure label on the diagnostic aliquot.
Evidence that argues against this (1)
- If the diagnostic aliquot were in fact collected after a first dose, the 745 would be an on-drug count and the pair would not be a clean before/after - but the paper's text explicitly places the 46-site measurement at diagnosis. Janssens DH et al. 2024
the interval-set dependence that bounds every count in this beat
the 46-site figure is the paper's own for the diagnostic, pre-drug aliquot
independent confirmation from the paper text that 148752 is at diagnosis
152985 was sampled during revumenib and calls zero promoters under every interval set