At promoter resolution the published N/C oncoprotein criterion calls 1043 promoters in fusion-negative CD34+ HSPCs - above the median across 27 KMT2A-rearranged patients - while the same-reagent controls in that sample call 66 and 11
Healthy blood stem cells have no fusion protein at all, yet the published test says hundreds of their gene promoters are fusion-bound when the test is run on promoters instead of on the regions it was designed for. The cutoff does not travel between region sizes.
The same thing, for a scientist
**The control.** CD34+ HSPCs carry **no KMT2A fusion**, so every promoter the criterion calls in them is a false positive. The sample is the one the paper calibrated its threshold on, deposited in **GSE159608** with a complete `KMT2A_N1/N2/C1/C2` AutoCUT&RUN set at the **same four antibody catalogue numbers** as the patient deposit (Millipore 05-764, CST 14689S, SantaCruz sc-374392, Millipore 05-765) [E1]. **Result.** On TSS +/- 1 kb promoters, the criterion calls **1043 promoters (7.98% of the 13067 measurable)**. The patient median is **947.00** and the patient minimum is **0** [E1]. **And it is not reagent noise.** In the same CD34+ sample, holding the epitope fixed and swapping the antibody - N1 against N2 - calls **66** promoters, and C1 against C2 calls **11** [E1]. **The N-versus-C asymmetry is 15.80x and 94.82x larger than the reagent-swap asymmetry**, in a cell with no fusion. Something systematic separates the N-terminal from the C-terminal signal in wild-type chromatin. **THE COUNTER-READING, WHICH IS THE MOST LIKELY ONE AND IS NOT A CRITICISM OF THE PAPER.** The authors calibrated this exact criterion on this exact sample to yield **3** false positives, on their own peak intervals [E2]. This beat gets 1043 on promoters. **The whole of that difference is the interval definition**, and the economical conclusion is that 1.75 is a threshold for peaks and does not carry to promoters - not that the published measurement is wrong. What this entry establishes is the size of the carry-over error, which was not previously on record: roughly two and a half orders of magnitude.
Genome builds were never mixed
GSE159608's fragment BEDs are hg19; GSE252378's are hg38. The CD34+ libraries were counted against an hg19 promoter set built by the same script and the same rule as the hg38 set (TSS +/- 1 kb, UCSC refGene, longest NM_ per symbol: 19392 hg19 promoters against 19404 hg38). No coordinate crosses builds anywhere in this analysis, and no per-promoter join between the CD34+ sample and the patients is performed - only counts and amplitudes are compared.
Library depths, so the comparison is not power
CD34+ totals: IgG 5,915,920, KMT2A_C1 9,236,259, KMT2A_C2 8,368,173, KMT2A_N1 7,139,881, KMT2A_N2 6,987,396 fragments. These are the same order as the patient libraries, so the false-positive count is not a product of unusual depth in the control.
| script | analyses/q144-unbounded-magnitude/run.py |
|---|---|
| status | run |
| result summary | CD34+ fusion-negative control: 1043 called promoters (7.98%), N1vN2 66, C1vC2 11; patient median 947.00. |
cd34
{
"accession": "GSE159608",
"genome_build": "hg19",
"promoters_total": 19392,
"promoters_measurable_at_20": 13067,
"library_totals": {
"IgG": 5915920,
"KMT2A_C1": 9236259,
"KMT2A_C2": 8368173,
"KMT2A_N1": 7139881,
"KMT2A_N2": 6987396
},
"gsms": {
"IgG": "GSM5139613",
"KMT2A_C1": "GSM5139616",
"KMT2A_C2": "GSM5139617",
"KMT2A_N1": "GSM5139618",
"KMT2A_N2": "GSM5139619"
},
"A_amplitude": 0.5703242203621195,
"p99_minus_p1": 3.1517676547288427,
"n_called_promoters": 1043,
"pct_called": 7.981939236243973,
"n_called_N1_over_N2": 66,
"n_called_C1_over_C2": 11,
"patient_median_n_called": 947,
"patient_min_n_called": 0,
"patient_A_median": 1.0011176709745866,
"note": "CD34+ HSPCs carry NO KMT2A fusion, so every promoter this criterion calls here is a false positive. Same four antibody catalogue numbers as GSE252378 (Millipore 05-764, CST 14689S/D2M7U, SantaCruz sc-374392/H10, Millipore 05-765). hg19 throughout, against an hg19 promoter set built by the same script; no cross-build coordinate is used anywhere."
}- Any oncoprotein-target gene list re-derived at promoter resolution from this criterion carries a false-positive fraction that a fusion-negative control says is large, and gene lists are what pharmacodynamic panels are built from.
- At what interval width does the CD34+ false-positive count fall to the published 3?
- Are the CD34+ calls a coherent gene set, or the top of the coverage distribution? That is the difference between fragment biology and statistical power.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
In fusion-negative CD34+ HSPCs (GSE159608, GSM5139616-GSM5139619), the published N/C criterion applied to TSS +/- 1 kb hg19 promoters calls 1043 promoters, against 66 for N1-vs-N2 and 11 for C1-vs-C2 in the same sample; the median across 27 KMT2A-rearranged patients is 947.00.
GSE159608 - Automated CUT&Tag profiling of chromatin heterogeneity in mixed-lineage leukemia (419 GSMs; CD34+ HSPC KMT2A N1/N2/C1/C2 AutoCUT&RUN, hg19) · Janssens DH, Henikoff S (Fred Hutchinson Cancer Center) · NCBI Gene Expression Omnibus 2021What this evidence cannot show
One donor, one CD34+ sample, one interval width, hg19. The patient comparison is a comparison of counts across two accessions and two builds, never a per-promoter join. No spike-in in either deposit. The measurable promoter universes differ between the two arms (13067 in CD34+ at the same n>=20 rule, 4646 in the patient intersection), so the counts are not normalised to a common universe and the percentages should be read before the raw counts.
Evidence that argues against this (1)
- The patient universe is an INTERSECTION across all samples and so is much smaller than the CD34+ universe, which is a single sample. A smaller universe can only produce fewer calls, so part of the gap between the CD34+ count and the patient counts is universe size rather than specificity. The percentage figures are the fair comparison and they narrow the gap; they do not close it. Janssens DH et al. 2021
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The paper set this threshold to produce 3 false-positive sites in the CD34+ oncoprotein-negative control sample, on its own called intervals.
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 paper's Results text. The authors' interval definition and any additional filters applied before scoring are not reproduced here.
Evidence that argues against this (1)
- If the authors' intervals are pre-selected for high coverage, their 3 false positives were achieved at much higher counts per interval than a promoter carries, so the comparison to 3 is a comparison across statistical power as well as across interval definition. Both differences point the same way and neither is separated here. Janssens DH et al. 2024
the control whose location that entry establishes
the fusion-negative N/C signal this hypothesis predicts
an independent control pointing the same way as the voiding clause
the recalibration this result makes necessary
CD34+ HSPCs as the fusion-negative control
in the fusion-negative control the same-protein pairs are clean and N-versus-C is not
the false-positive rate that contaminates both sets in every pair
second independent demonstration that the criterion fires without an oncoprotein