A promoter neighbour goes missing two ways - absent from the table, or present under a different one of its own TSSs - and both return the same clean negative

{
  "name": "absent from the table",
  "cause": "one NM_-derived row per symbol",
  "worked_case": "AFF1-AS1",
  "remedy": "audit against the full annotation (Q153)",
  "bounded_at": "7.2 percentage points, from Q153's 14.37% within-table vs 21.60% full-annotation censuses"
}

{
  "name": "one row per symbol, where the symbol has several non-overlapping TSSs",
  "cause": "the table schema keeps one TSS per gene symbol",
  "worked_case": "C4orf36, three current curated TSSs spanning 43,782 bp, table keeps the distal one",
  "remedy": "keep all of a symbol's TSSs, or state which one and why; refreshing the annotation does NOT help",
  "bounded_at": "14.69% of symbols (2,851 of 19,404), median span 21,291 bp, 1,939 with an alternate TSS >10 kb away (Q156)"
}

{
  "name": "one symbol, two gene models",
  "cause": "HGNC-anchored and NCBI-anchored models coexist under one display name",
  "worked_case": "C4orf36 ENSG00000163633 vs ENSG00000285458, TSSs 43,365 bp apart",
  "remedy": "state which source resolved the symbol to coordinates",
  "bounded_at": "unmeasured; this repository currently resolves from a fixed table, so it is prophylaxis"
}

[
  {
    "quantity": "pct_symbols_with_multiple_non_overlapping_tss",
    "subject": "rows of this project's promoter table whose symbol has more than one non-overlapping current TSS",
    "value": 14.69,
    "units": "percent of rows",
    "n": 19404,
    "universe": "promoters.tsv joined against the current UCSC hg38 ncbiRefSeqCurated, NM_-only comparison (Q156)",
    "null": 1.83,
    "note": "the null is the genuine-displacement rate from the same audit; the schema blind spot is 8x the drift blind spot"
  },
  {
    "quantity": "pct_rows_displaced",
    "subject": "rows whose 2 kb window is disjoint from every current window for the same symbol",
    "value": 1.83,
    "units": "percent of rows",
    "n": 19404,
    "universe": "same audit, NM_-only; 355 rows, 81.63% TSS-identical, 0 moved chromosome (Q156)",
    "null": 7.2,
    "note": "the null is Q153's blind spot 1 (neighbours with no row) at 7.2 percentage points; drift is four times smaller"
  },
  {
    "quantity": "multi_tss_span_median",
    "subject": "median distance between the most distal and most proximal current TSS, among multi-TSS symbols",
    "value": 21291,
    "units": "bp",
    "n": 2851,
    "universe": "the 2,851 multi-TSS symbols of the promoter table (Q156)",
    "null": 2000,
    "note": "the null is the window width: the median multi-TSS gene's own start sites are ten window-widths apart"
  },
  {
    "quantity": "tss_separation_between_two_ensembl_models_of_one_symbol",
    "subject": "ENSG00000163633 against ENSG00000285458, both displayed as C4orf36",
    "value": 43365,
    "units": "bp",
    "n": 2,
    "universe": "Ensembl REST GRCh38 lookup/id, minus-strand TSSs 86,892,813 and 86,936,178",
    "null": 0,
    "note": "0 is what two ids for the same gene should differ by if a symbol-keyed lookup is to be safe"
  }
]

{
  "name": "annotation drift (WITHDRAWN as the second blind spot)",
  "why_withdrawn": "This entry first named drift as blind spot 2 on the strength of a 42,310 bp C4orf36 displacement that does not exist. Q156's whole-table audit measures genuine drift at 1.83% of rows, four times SMALLER than Q153's blind spot 1, with 81.63% of rows TSS-identical and no row moved chromosome. Drift is real, bounded and small; it is not the second blind spot.",
  "bounded_at": "1.83% of rows (355 of 19,404), Q156"
}