start / graph / ana-q144-two-antibodies-against-the-same-end-of-kmt2a-call-as-many-oncoprotein-target-promoters-as-the-real-n-over-c-comparison-does

AnalysisstrongQ144-unbounded-magnitude

In 18 of 27 patient samples, comparing two antibodies raised against the SAME end of KMT2A calls at least as many oncoprotein-target promoters as the real N-over-C comparison does, and swapping the antibody pair moves a sample's own call count by a median of 2.57x

The test for fusion binding compares antibodies against the front and back of a protein. If you instead compare two antibodies against the same end, where there is nothing to find, the test still fires - in many patients as often as the real comparison does.

The same thing, for a scientist

**The control.** `N1` and `N2` are two antibodies against the KMT2A N-terminus; `C1` and `C2` are two against the C-terminus. Neither pair contains a protein species the other lacks, so under the published criterion each pair should call nothing [E1]. **Measured in all 27 samples at TSS +/- 1 kb promoters.** Median worst-case same-protein call count **1385.00**, against a median real N-over-C count of **947.00**; median ratio of real to control **0.50**; and the control equals or exceeds the real signal in **18 of 27 samples** [E1]. **And the real call count itself moves with the reagent.** Scoring each antibody pair on its own (N1/C1 versus N2/C2) changes a single sample's call count by a median factor of **2.57** [E2]. Across samples the two estimates still rank together - Spearman **0.397**, Pearson on log(1+n) **0.679** - so the ordering of patients by oncoprotein burden is reproducible while its **value** is not [E2]. **Which is exactly why this entry does not overturn the magnitude result.** The between-patient spread of the call count is **20.01x** at the 10th-to-90th percentile and **103.73x** from the largest to the smallest non-zero count, against a within-patient reagent-swap factor of **2.57**. The signal is between one and two orders of magnitude larger than its own reagent noise. What the reagent noise forbids is reading any individual sample's count as a quantity; what it permits is the comparison between samples.

Go deeper

Per-sample same-protein calls

sample	N1>N2	N2>N1	C1>C2	C2>C1	N>C (real)
TB040576	755	1348	582	1808	2282
p247	547	50	213	804	2095
R3652	7	3651	1352	402	2063
p160	1435	133	1571	316	2060
p279	1551	72	827	628	2045
p454	918	77	250	1635	2000
p318	283	295	283	1294	1894
R3595	3	4219	429	342	1233
TB000694	174	487	817	387	1207
TB986292	149	759	630	172	1151
A93692	224	1500	77	604	1087
A89691	49	277	22	101	1069
TB051939	297	2415	1097	351	1048
p186	219	414	171	2126	947
TB971992	275	1197	795	809	893
TB995291	590	1775	389	461	890
p214	1334	90	484	880	767
p179	622	61	250	1514	712
TB920488	901	1019	756	102	429
A82485	322	989	827	316	400
148752	0	4582	0	3209	345
A80956	25	802	104	50	332
A95941	644	1264	228	1385	237
A93949	1377	581	15	3375	147
A70498	42	206	74	15	37
A73051	46	152	1083	2	22
152985	0	2	2	0	0

The direction of the reagent asymmetry

The N2-over-N1 direction - rabbit CST D2M7U over mouse Millipore 05-764 - calls more promoters than N1-over-N2 in most samples. A directional excess between two antibodies against one epitope region, differing in host species, is the signature of an efficiency or background difference rather than of a protein-species difference. C1 and C2 are both mouse and are the cleaner of the two pairs, which is consistent with the host term being real.

What this does NOT license

It does not license discarding the published oncoprotein-target list. The published calls are made on the authors' own intervals with a threshold calibrated there, and this control was run at a different interval scale. What it licenses is a refusal to re-derive oncoprotein-target sets at promoter resolution from the bare 1.75 threshold without running this control first - which costs nothing, because the libraries needed are already in the same deposit.

Facts and figures

script	analyses/q144-unbounded-magnitude/run.py
status	run
result summary	Same-protein antibody pairs call a median worst-case 1385.00 promoters vs 947.00 for real N/C; control >= signal in 18 of 27 samples; reagent swap moves a sample's own count 2.57x.

per sample

per sample — the numbers behind the chart

{
  "148752": {
    "N1_over_N2": 0,
    "N2_over_N1": 4582,
    "C1_over_C2": 0,
    "C2_over_C1": 3209,
    "N_over_C": 345
  },
  "152985": {
    "N1_over_N2": 0,
    "N2_over_N1": 2,
    "C1_over_C2": 2,
    "C2_over_C1": 0,
    "N_over_C": 0
  },
  "A70498": {
    "N1_over_N2": 42,
    "N2_over_N1": 206,
    "C1_over_C2": 74,
    "C2_over_C1": 15,
    "N_over_C": 37
  },
  "A73051": {
    "N1_over_N2": 46,
    "N2_over_N1": 152,
    "C1_over_C2": 1083,
    "C2_over_C1": 2,
    "N_over_C": 22
  },
  "A80956": {
    "N1_over_N2": 25,
    "N2_over_N1": 802,
    "C1_over_C2": 104,
    "C2_over_C1": 50,
    "N_over_C": 332
  },
  "A82485": {
    "N1_over_N2": 322,
    "N2_over_N1": 989,
    "C1_over_C2": 827,
    "C2_over_C1": 316,
    "N_over_C": 400
  },
  "A89691": {
    "N1_over_N2": 49,
    "N2_over_N1": 277,
    "C1_over_C2": 22,
    "C2_over_C1": 101,
    "N_over_C": 1069
  },
  "A93692": {
    "N1_over_N2": 224,
    "N2_over_N1": 1500,
    "C1_over_C2": 77,
    "C2_over_C1": 604,
    "N_over_C": 1087
  },
  "A93949": {
    "N1_over_N2": 1377,
    "N2_over_N1": 581,
    "C1_over_C2": 15,
    "C2_over_C1": 3375,
    "N_over_C": 147
  },
  "A95941": {
    "N1_over_N2": 644,
    "N2_over_N1": 1264,
    "C1_over_C2": 228,
    "C2_over_C1": 1385,
    "N_over_C": 237
  },
  "R3595": {
    "N1_over_N2": 3,
    "N2_over_N1": 4219,
    "C1_over_C2": 429,
    "C2_over_C1": 342,
    "N_over_C": 1233
  },
  "R3652": {
    "N1_over_N2": 7,
    "N2_over_N1": 3651,
    "C1_over_C2": 1352,
    "C2_over_C1": 402,
    "N_over_C": 2063
  },
  "TB000694": {
    "N1_over_N2": 174,
    "N2_over_N1": 487,
    "C1_over_C2": 817,
    "C2_over_C1": 387,
    "N_over_C": 1207
  },
  "TB040576": {
    "N1_over_N2": 755,
    "N2_over_N1": 1348,
    "C1_over_C2": 582,
    "C2_over_C1": 1808,
    "N_over_C": 2282
  },
  "TB051939": {
    "N1_over_N2": 297,
    "N2_over_N1": 2415,
    "C1_over_C2": 1097,
    "C2_over_C1": 351,
    "N_over_C": 1048
  },
  "TB920488": {
    "N1_over_N2": 901,
    "N2_over_N1": 1019,
    "C1_over_C2": 756,
    "C2_over_C1": 102,
    "N_over_C": 429
  },
  "TB971992": {
    "N1_over_N2": 275,
    "N2_over_N1": 1197,
    "C1_over_C2": 795,
    "C2_over_C1": 809,
    "N_over_C": 893
  },
  "TB986292": {
    "N1_over_N2": 149,
    "N2_over_N1": 759,
    "C1_over_C2": 630,
    "C2_over_C1": 172,
    "N_over_C": 1151
  },
  "TB995291": {
    "N1_over_N2": 590,
    "N2_over_N1": 1775,
    "C1_over_C2": 389,
    "C2_over_C1": 461,
    "N_over_C": 890
  },
  "p160": {
    "N1_over_N2": 1435,
    "N2_over_N1": 133,
    "C1_over_C2": 1571,
    "C2_over_C1": 316,
    "N_over_C": 2060
  },
  "p179": {
    "N1_over_N2": 622,
    "N2_over_N1": 61,
    "C1_over_C2": 250,
    "C2_over_C1": 1514,
    "N_over_C": 712
  },
  "p186": {
    "N1_over_N2": 219,
    "N2_over_N1": 414,
    "C1_over_C2": 171,
    "C2_over_C1": 2126,
    "N_over_C": 947
  },
  "p214": {
    "N1_over_N2": 1334,
    "N2_over_N1": 90,
    "C1_over_C2": 484,
    "C2_over_C1": 880,
    "N_over_C": 767
  },
  "p247": {
    "N1_over_N2": 547,
    "N2_over_N1": 50,
    "C1_over_C2": 213,
    "C2_over_C1": 804,
    "N_over_C": 2095
  },
  "p279": {
    "N1_over_N2": 1551,
    "N2_over_N1": 72,
    "C1_over_C2": 827,
    "C2_over_C1": 628,
    "N_over_C": 2045
  },
  "p318": {
    "N1_over_N2": 283,
    "N2_over_N1": 295,
    "C1_over_C2": 283,
    "C2_over_C1": 1294,
    "N_over_C": 1894
  },
  "p454": {
    "N1_over_N2": 918,
    "N2_over_N1": 77,
    "C1_over_C2": 250,
    "C2_over_C1": 1635,
    "N_over_C": 2000
  }
}

call count reliability

{
  "spearman_N1C1_vs_N2C2_across_samples": 0.39658067932662394,
  "pearson_log1p": 0.6791302165375804,
  "median_within_sample_ratio": 2.5719557195571956,
  "note": "Each antibody pair scored on its own under the published criterion.  The across-sample correlation of the two is the reliability of the call count as a between-sample quantity."
}

Still open

Would a same-host pair of N-terminal antibodies give a clean control? No deposit in this corpus's reach contains one for KMT2A.

Every source behind this entry (2)

E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
In GSE252378, comparing the two N-terminal KMT2A antibodies against each other, or the two C-terminal antibodies against each other, under the published N/C criterion at TSS +/- 1 kb promoters calls a median worst-case 1385.00 promoters against a median 947.00 for the real N-over-C comparison, with the control equalling or exceeding the signal in 18 of 27 samples.
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 2024
What this evidence cannot show
Promoters, not the authors' peak intervals. The two pairs are not equally good controls: N1 is mouse and N2 rabbit, so that comparison carries a host-species term the pooled N-over-C partly averages out, while C1 and C2 are both mouse. Within-sample shares, no spike-in.
Evidence that argues against this (1)
- A reagent pair is arguably a harsher control than any ratio criterion could survive: CUT&RUN efficiency is antibody-specific, so two antibodies against one epitope region do report different amounts of the same protein and a ratio test will find that. On that reading the control measures a real property of antibodies rather than a defect of the criterion - which is true, and is precisely why a criterion moved to a new interval scale needs its own calibration rather than an inherited threshold. Janssens DH et al. 2024
E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
Scoring each antibody pair separately under the same criterion changes a sample's own call count by a median factor of 2.57, while the two estimates still rank samples together at Spearman 0.397 and Pearson 0.679 on log(1+n).
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 2024
What this evidence cannot show
Each antibody pair uses half the fragments of the pooled comparison, so part of the within-sample factor is lower power rather than reagent difference; the pooled count is not the average of the two and is not expected to be. The ranking statistics are across 27 samples.
Evidence that argues against this (1)
- A median 2.6-fold reagent swing means no single sample's call count is a measurement of anything in absolute terms, and any future entry quoting one patient's oncoprotein-target count as a number should be read as carrying at least that much uncertainty before any biological interpretation. Janssens DH et al. 2024

Connected to

depends on →GSE252378 fragment BEDs: a complete KMT2A N-terminal / C-terminal antibody set in 27 patient samples across six fusion partners, deposited and until now unopened by this corpus
the two-antibody-per-arm design that makes this control possible
supports →The unbounded oncoprotein score spans 1.73-12.32 log2 units within a sample - it is not the compressed statistic q142 was stuck with - and the pre-registered magnitude verdict is STILL void, because the score's amplitude is no larger than the disagreement between two antibodies against the same end of KMT2A
the criterion-space version of the pre-registered voiding clause
related to →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
in the fusion-negative control the same-protein pairs are clean and N-versus-C is not
supports →Scored in the published criterion's own currency, the number of oncoprotein-target promoters runs 0 to 2282 across 27 KMT2A-rearranged patient samples - 20.01x between the 10th and 90th percentiles - and the spread widens to 61.10x after depth matching
sizes the reagent noise the magnitude spread has to beat, and it beats it
related to →KMT2A / MLL1 — the gene that breaks
four KMT2A antibodies, two per terminus
← supportsThe excess overlap survives a structure-respecting null in both beats - permutation p = 0.0005 in all six cells - but every published excess figure shrinks 2.4- to 4-fold
replicated at a different depth matching: 62% and 61% of the real call count