The per-library 'cannot be occupied' floor spans 0.00x to 4.01x an average promoter, and one core gene's AFF1-exclusivity rests entirely on it
also called q152 floor audit
The occupancy test uses eight genes that should never be switched on in leukaemia as a per-sample noise floor. In one sample that floor reads four times an average gene, so in that sample almost nothing can pass. One of the seventeen 'AFF1-only' genes is only AFF1-only because of that one sample's high floor.
The same thing, for a scientist
The q147/q151 rule requires a promoter to exceed the per-library MAXIMUM of eight width-matched cannot-be-occupied promoters (ALB, GFAP, INS, KRT14, MYH7, MYOD1, NEUROD1, RHO). Across the 54 KMT2A-N libraries that floor spans 0.00x to 4.01x an average promoter in replicate 1 and 0.00x to 1.08x in replicate 2. A 4.01x floor is above the 2.0x gate, so in that library (A80956, the 1.86% pass-rate sample) the effective threshold is doubled. Exactly one of the 17 core promoters - NOP58 - has a non-AFF1 sample (A80956) that clears 2.0x in BOTH replicates (2.66x and 7.14x) and is rejected only by the replicate-1 floor of 4.005x. NOP58's AFF1-exclusivity is a property of that one library's floor, not of its occupancy.
The floor is doing more work than the threshold in the shallow libraries
A per-library floor is the right idea: it stops a shallow or poorly enriched library calling everything occupied. But it is a maximum of eight, so it is an extreme-value statistic and in a library where one of the eight negative promoters happens to catch reads it moves a long way. Here it moves to 4.01x - twice the nominal gate 1.
The consequence is asymmetric. A high floor makes a library call nothing, which makes every promoter look exclusive to whichever arm that library is not in. A80956 is a KMT2A-AFDN sample, so its high floor systematically inflates the apparent exclusivity of the AFF1 arm.
Named exclusion, counted
A filter must report what it removed. Of the 17 core promoters, 1 (NOP58) has at least one non-AFF1 sample removed from the comparator by the floor alone; 16 have none. Of the 19 non-AFF1 samples, the floor is above 2.0x in one (A80956, 4.005x in replicate 1). No core promoter is affected by the replicate-2 floors, all of which are below 1.08x 1.
| script | analyses/q152-the-aff1-only-recurrent-promoter-core/step4.py |
|---|---|
| status | run |
| result summary | Floor spans 0.00-4.01x (rep1) and 0.00-1.08x (rep2). One of 17 core genes (NOP58) is AFF1-exclusive only because of A80956's 4.005x replicate-1 floor. |
floor promoters
[ "ALB", "GFAP", "INS", "KRT14", "MYH7", "MYOD1", "NEUROD1", "RHO" ]
measurements
[
{
"quantity": "max_per_library_floor",
"subject": "the per-library maximum of eight width-matched cannot-be-occupied promoters, KMT2A-N replicate 1",
"value": 4.005,
"units": "x an average promoter in that library",
"n": 27,
"universe": "27 GSE252378 samples, within-sample shares over 19,404 refGene 2 kb promoters",
"null": 1,
"note": "sample A80956; range across samples 0.00-4.01x; the nominal occupancy gate is 2.0x, so this library's effective gate is doubled"
},
{
"quantity": "n_core_genes_rescued_by_the_floor",
"subject": "core promoters whose AFF1-exclusivity depends on a floor-only exclusion of a non-AFF1 sample",
"value": 1,
"units": "promoters",
"n": 17,
"universe": "the 17 AFF1-only recurrent core promoters",
"null": null,
"note": "NOP58 only; the other 16 have no non-AFF1 sample clearing 2.0x in both replicates"
}
]- Should the floor be the maximum of eight, the median, or a quantile? The maximum is an extreme-value statistic on eight draws and is why one library's floor reaches 4.01x.
- Do the eight floor promoters have any residual accessibility in blasts, or is the 4.01x reading in A80956 traceable to one promoter? The per-promoter decomposition was not run.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
Across the 27 GSE252378 samples the per-library maximum of the eight width-matched cannot-be-occupied promoters ranges 0.00-4.01x an average promoter in KMT2A-N replicate 1 and 0.00-1.08x in replicate 2; NOP58 reads 2.66x and 7.14x in the two KMT2A-AFDN sample A80956 replicates and is called unoccupied there solely because the replicate-1 floor is 4.005x.
Q152 per-library floor audit and floor-only exclusions · Iqbal Lab oncofusion knowledge graph, beat Q152 · this project 2026What this evidence cannot show
The eight floor promoters are the set this project has used since q145; the audit does not ask whether a different eight would behave better, only what this eight does.
Evidence that argues against this (1)
- The high floor may be correct rather than an artefact: A80956 is also the lowest-pass-rate library in the deposit (1.86%), which is consistent with a genuinely poor enrichment where a high noise floor is the right call. In that reading NOP58's 2.66x in that library is noise and the floor did its job. The two readings cannot be separated without a spike-in, which this deposit does not have. NCBI Gene Expression Omnibus 2024
NOP58's inverted margin of 0.75 is the floor's doing
audits the per-library floor that q151's rule uses
NOP58 fails the patient-collapse check and the floor check independently
the promoter-scale half of the same audit; the contrast between the two scales is the point