Thirteen of q152's seventeen AFF1-only core promoters are absent from the deposit paper's whole target-gene table, and the four that are present are AFF1-exclusive there too
also called q152 core cross-checked against Supplementary Data 2 · the core's four survivors
An earlier round of this project picked out seventeen genes that seemed to be switched on by the fusion only in the AFF1 leukaemias. When we look those seventeen up in the list the original scientists published, thirteen are not in it at all - their method never counted them as targets anywhere, in any patient. The four that are there are indeed AFF1-only.
The same thing, for a scientist
q152 nominated 17 promoters as an AFF1-only recurrent core (occupied in >=3 of 8 KMT2A::AFF1 samples, 0 of 19 non-AFF1) and found the count indistinguishable from a label-permutation null [E2]. Cross-checking those 17 against Janssens et al.'s published 1,091-gene oncoprotein-target table [E1]: **13 appear nowhere in it** - AMZ2, CCP110, GBA, MAGI3, MED28, MTFR1, NOP58, PBLD, POU2AF1, SPRED1, SYNE3, TBX21, UHRF1BP1L - meaning the authors' peak-plus-N/C pipeline never called them a target in **any** of 37 samples including cell lines. The four that are present are: **BLNK** (p179, p247, p279, p454 - 4 AFF1, 0 non-AFF1, and a member of the 64-gene AFF1-exclusive class), **WT1** (p179, p247 - 2 AFF1, 0 non-AFF1), **EPHA7** (p279 only), and **MATR3** (p179, p247, p279 AFF1 plus p214, which is KMT2A-ENL, so not exclusive) [E1]. **Three of the four survivors are AFF1-exclusive under the stricter instrument.** So q152's core is not wrong about direction where it can be checked; it is mostly below the threshold of the instrument that generated the data.
The table
| q152 core gene | in the authors' 1,091-gene table | AFF1 samples called | non-AFF1 samples called |
|---|---|---|---|
| BLNK | yes | p179, p247, p279, p454 (4) | none |
| MATR3 | yes | p179, p247, p279 (3) | p214 (KMT2A-ENL) |
| WT1 | yes | p179, p247 (2) | none |
| EPHA7 | yes | p279 (1) | none |
| AMZ2, CCP110, GBA, MAGI3, MED28, MTFR1, NOP58, PBLD, POU2AF1, SPRED1, SYNE3, TBX21, UHRF1BP1L | no | - | - |
BLNK is one of the 64 members of the AFF1-exclusive recurrent class that sits far above its permutation null in the authors' table.
Which way this cuts, stated both ways
For q152's core: where the two instruments overlap they agree on direction, and BLNK in particular is AFF1-exclusive under both. q152's negative was about the count of such promoters against a null, not about any member being wrong, and this does not change that.
Against q152's core: 13 of 17 being invisible to the authors' caller is not a small discrepancy. It means the core is largely built from signal that does not survive IgG subtraction and an FDR-corrected N/C test. q152 already recorded that nine of the seventeen have a non-AFF1 sample within 10% of the threshold and that NOP58 rests entirely on the per-library floor 2; this is a third, independent reason to treat the 13 as weak.
What it does not show: absence from the authors' table is not proof a gene is not a target. Their master list is 1,692 intervals for the whole study, built from peaks called on pooled libraries; a real but modest site in one or two samples can fail to be nominated at all. This is a bound on what their pipeline reports, not on what is on chromatin.
| script | analyses/q154-gene-scale-occupancy-null/authors_null2.py |
|---|---|
| status | run |
| result summary | 4 of 17 q152 core genes appear in the authors' 1,091-gene target table; 3 of those 4 are AFF1-exclusive there; MATR3 is not (p214 is KMT2A-ENL). |
inputs
[ "Janssens 2024 Supplementary Data 2 (41467_2024_53399_MOESM4_ESM.xlsx)", "analyses/q152-the-aff1-only-recurrent-promoter-core/results.json (step1.core)" ]
present and aff1 exclusive
[ "BLNK", "WT1", "EPHA7" ]
present but not exclusive
[ "MATR3" ]
absent from the authors table
[ "AMZ2", "CCP110", "GBA", "MAGI3", "MED28", "MTFR1", "NOP58", "PBLD", "POU2AF1", "SPRED1", "SYNE3", "TBX21", "UHRF1BP1L" ]
measurements
[
{
"quantity": "fraction_of_a_nominated_core_visible_to_the_source_instrument",
"subject": "q152's 17 AFF1-only recurrent core promoters, looked up in the deposit paper's own oncoprotein-target table",
"value": 4,
"units": "genes of 17",
"n": 17,
"universe": "Janssens 2024 Supplementary Data 2, 1,091 oncoprotein-target genes called across 37 samples",
"null": null,
"note": "3 of the 4 present are AFF1-exclusive in that table; MATR3 is also called in the KMT2A-ENL sample p214"
}
]- If the 13 are artefacts of a permissive rule, any target list this project circulates from that rule carries roughly a three-in-four chance per entry of not being reproducible by the pipeline that generated the data.
- Are the 13 absent genes absent because their signal is below the authors' FDR-corrected N/C threshold, or because no peak was ever nominated at their promoters? Those are different failures and the master list of 1,692 intervals would separate them.
- BLNK is AFF1-exclusive under both instruments and is a B-cell linker protein in a B-lineage-restricted fusion. Is BLNK a better candidate for the partner-specific address than IRX1, which rests on two patients?
- Would matching this project's rule to the authors' call volume - about 125 genes per sample rather than about 2,086 promoters - recover the 13, or eliminate them?
an independent reason to distrust 13 of the 17: the deposit paper's own caller nominates them in no sample at all
BLNK is AFF1-exclusive under both instruments - 4 of 8 AFF1, 0 of 18 non-AFF1 in the authors' own table
EPHA7 is present in the authors' table but in only one sample, p279
same supplementary table