Exact GO:0006915 human, mouse, and zebrafish snapshot

Human and mouse were fetched from the QuickGO annotation API on 2026-09-30 with
fetch_go0006915.py, using goId=GO:0006915,
goUsage=exact, and taxonId set to either human 9606 or mouse 10090.
Zebrafish was added on 2026-10-05 with taxonId set to 7955.
Descendant annotations were deliberately excluded so this is only the broad
top-level apoptotic process row set.

Generated files

To refresh the standing human and mouse snapshot from the repository root:

uv run projects/APOPTOSIS/fetch_go0006915.py

To refresh the zebrafish extension:

uv run projects/APOPTOSIS/fetch_go0006915.py --taxon zebrafish

Snapshot size

Rows are not unique biological assertions. The human TP53 block alone has
103 rows because broad InterPro IEA rows are repeated over many UniProt
accessions and isoforms; the symbol rollups are therefore a better worklist
view than the raw rows.

Evidence-code mix

Human exact rows are dominated by electronic sources, chiefly InterPro and
UniProt IEA imports. Mouse has a larger experimental fraction because MGI
contributes many IMP rows from survival/death-readout genetics; those are the
rows most likely to need an over-annotation check before accepting a generic
GO:0006915 assertion.

Top assigners

Zebrafish InterPro2GO rows

The exact zebrafish slice is mostly not manual ZFIN curation. InterPro contributes
169 of the 253 exact GO:0006915 rows, and six high-count InterPro2GO mappings
from the largest-symbol worklist account for 127 rows on their own:

These are likely InterPro2GO mapping issues, so the highest-yield fix is upstream
signature review rather than one zebrafish gene at a time. IPR028754, which
maps the PAK5 kinase subfamily straight to exact apoptosis, and IPR002117,
which maps the whole p53/p63/p73 family to exact apoptosis, are especially strong
candidates to challenge.

Largest symbol rollups

ZFIN experimental row triage

The ZFIN-authored exact GO:0006915 slice is small: 21 rows, all without
annotation extensions, collapse to seven (symbol, reference) packages.
The first-pass curation estimate is metadata-based: it uses GOA evidence
metadata, with_from reagents, gene identity, and cached or abstract-level PMID
checks rather than full-text review of all seven papers. Only
timm50/PMID:15044455 has a cached publication record in this checkout.

This is a row count, not a unique-biology count: the tp53 package alone
contributes eight UniProt rows for the same ZFIN genotype or morpholino evidence,
and casp3a, hpdl, dmrt2a, and timm50 are also repeated over accessions.

The seven likely over-annotation rows split into two evidence shapes. hpdl and
rpl11 are IMP packages where perturbing a mitochondrial
4-hydroxyphenylpyruvate dioxygenase-like protein or a ribosomal protein changes
a downstream TUNEL, acridine-orange, caspase, or generic cell-death signal; the
first rpl11 row even shares ZFIN:ZDB-MRPHLNO-070126-7 with the tp53 package,
consistent with p53-dependent ribosomal-stress apoptosis. The dmrt2a rows are
different because they are IDA rows with no with_from reagent, but a direct
top-level BP assertion on a DM-domain transcription factor is still broad enough
to need ZFIN review before a DANRE review accepts exact apoptotic process.

The 14 likely pushdown rows are qualitatively different. casp3a should move
toward execution-phase apoptosis, tp53 toward p53-mediated apoptotic signaling
or its regulation, and timm50 toward the intrinsic branch, release of cytochrome
c from mitochondria, or negative regulation of cytochrome c release. The
single-row ptmab package is plausibly apoptosome/caspase-3 regulation, but it
needs full-text checking before assigning a directional regulation term.

Already reviewed APOPTOSIS targets

The current Priority 1 human review set still covers 132 exact top-level rows:
FAS 18, CASP8 14, DIABLO 14, BCL2L1 11, DFFB 10, APAF1 8,
CASP9 8, DFFA 6, CASP3 5, FASLG 5, BAX 4, BCL2 4, CASP7 4,
CASP6 3, PMAIP1 3, AIFM1 2, BAK1 2, BCL2L11 2, CASP10 2,
BBC3 1, CFLAR 1, FADD 1, TNFRSF10A 1, TNFRSF10B 1, TNFRSF1A 1,
and TNFSF10 1.

The mouse over-annotation slice covered three symbols with exact top-level rows:
Bcl2 9, Casp3 9, and Tnfrsf1a 1. The broader mouse rollup points to
Casp9, Bad, Dap3, Ripk1, Casp8, Col6a1, Diablo, and Bcl2l1 as
the next reviewed-gene comparators if this project continues the mouse branch.