Oncofusion

start / graph / ana-q152-two-of-the-seventeen-core-promoter-windows-are-shared-with-a-neighbouring-tss-and-pbld-shares-38-percent-with-hnrnph3

Analysisstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q152-aff1-only-recurrent-core

Two of the seventeen core promoter windows are shared with a neighbouring TSS - PBLD overlaps HNRNPH3 by 38.2% and CCP110 overlaps GDE1 by 7.8%

also called q152 C4 shared-window check

Each gene's switch is measured as a two-kilobase stretch centred on where transcription starts. If a second gene starts nearby, the same stretch covers both and you cannot tell which gene the signal belongs to. Two of the seventeen have that problem.

The same thing, for a scientist

Pre-registered escape condition C4, run against the same 19,404-row refGene TSS table the promoter windows were cut from. Of the 17 core promoters, 15 contain no other refGene TSS within 2 kb. **PBLD**'s TSS is 1,236 bp from HNRNPH3's, both on the plus strand, so their 2 kb windows overlap **38.2%**. **CCP110**'s TSS is 1,843 bp from GDE1's on the minus strand, a **7.8%** overlap. Both are reported as unattributable and both are named. IRX1, checked as a control because of the IRX2/IRX2-DT precedent (215 bp apart, 89.2% window overlap), has no neighbouring TSS within 2 kb.

Go deeper
The named exclusion, counted
core promoterneighbourTSS distancestrandwindow overlap
PBLDHNRNPH31,236 bp+ (same)38.2%
CCP110GDE11,843 bp-7.8%
other 15none within 2 kb--0%

PBLD's 38.2% overlap with HNRNPH3 - a heterogeneous nuclear ribonucleoprotein, a far more highly expressed gene than PBLD - means PBLD's occupancy call cannot be attributed to PBLD. CCP110's 7.8% overlap is small enough to be unlikely to drive a call but is recorded 1.

Neither removal changes the beat's verdict, which is already negative; the exclusions are recorded because a filter must report what it removed even when the removal does not matter.

A symbol note on CCP110

CCP110 is not the current official symbol. HGNC's approved symbol for this gene is CP110 (centriolar coiled-coil protein 110); CCP110 is a previous symbol carried in the refGene table this project's promoter windows were built from. Any literature search on this gene must run CP110 as well as CCP110, per this project's standing rule that a previous official symbol is not an alias. The same caution applies to GBA, whose approved symbol is now GBA1.

Facts and figures
scriptanalyses/q152-the-aff1-only-recurrent-promoter-core/step4.py
statusrun
result summary2 of 17 core windows are shared; PBLD/HNRNPH3 38.2%, CCP110/GDE1 7.8%; the other 15 are clean.
measurements
[
  {
    "quantity": "n_core_promoters_with_a_shared_window",
    "subject": "the 17 AFF1-only recurrent core promoters",
    "value": 2,
    "units": "promoters",
    "n": 17,
    "universe": "19,404 representative refGene TSSs, one per gene symbol, 2 kb TSS-centred windows",
    "null": null,
    "note": "PBLD/HNRNPH3 at 1,236 bp (38.2% overlap) and CCP110/GDE1 at 1,843 bp (7.8%); alternative and non-coding TSSs are not in the table and were not searched"
  }
]
Still open
  1. Do any of the 15 clean core promoters have a divergent lncRNA TSS absent from the refGene representative table? The IRX2/IRX2-DT case was found only by name-based search and a systematic GENCODE-based check has never been run for this project's promoter set.
  2. ANSWERED CONCURRENTLY BY Q153, same cycle: 21.60% of the 19,404 windows contain another gene's hg38 refGene TSS within 1 kb and 14.37% contain one that is itself a row of the project's table. This open question closed the day it was written; the remaining part is whether a GENCODE-based check finds more, since q153 used refGene as this beat did.
Every source behind this entry (1)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

    Of the 17 AFF1-only recurrent core promoters, 15 contain no other refGene TSS within 2 kb; PBLD's TSS is 1,236 bp from HNRNPH3's on the same strand giving a 38.2% window overlap, and CCP110's TSS is 1,843 bp from GDE1's giving a 7.8% overlap; IRX1's window contains no other TSS within 2 kb.

    Q152 shared-promoter-window audit against the 19,404-row refGene TSS table · Iqbal Lab oncofusion knowledge graph, beat Q152 · this project 2026
    What this evidence cannot show

    The neighbour search uses only the 19,404 representative refGene TSSs in this project's promoter table, one per gene symbol. A gene with an alternative TSS closer to a core promoter, or a non-coding transcript absent from that table, would not be seen - which is precisely how the IRX2/IRX2-DT overlap was found only when a divergent lncRNA was looked for by name.

    Evidence that argues against this (1)
    • A concurrent beat (q153) censuses the same 19,404-row window table and reports that 14.37% of windows have a neighbouring TSS that is itself a row of that table within 1 kb (i.e. a >=50% window overlap), rising to 21.60% against full hg38 refGene. Both of this core's shared windows are BEYOND 1 kb (PBLD 1,236 bp, CCP110 1,843 bp), so at q153's >=50%-overlap criterion this core scores 0 of 17 against a 14.37% within-set base rate. This check is therefore less alarming than it reads, and q153's wider neighbour set is the one that matters: my search used only the one-row-per-symbol table, which q153 measures to understate the problem by 7.2 percentage points. Iqbal Lab oncofusion knowledge graph et al. 2026
Connected to