APLN (Q9ULZ1) — bioinformatics support for the annotation review

Three things in the APLN GOA record can be settled from sequence and from the committed
PAINT slice rather than asserted. Every script fetches live from UniProt, InterPro,
RNAcentral and RNAcentral's species views (responses cached under cache/, which is
disposable) and from interpro/panther/PTHR15953/PTHR15953-paint.tsv. Nothing below is
hardcoded; delete cache/ and re-run to regenerate every number.

uv run python cterm_conservation.py      # -> cterm_conservation.tsv
uv run python resolve_withfrom.py        # -> withfrom_resolved.tsv, source_entities.yaml
uv run python check_goa_reconciliation.py

1. Is the mature apelin peptide conserved, and is the rodent peptide the human peptide? (cterm_conservation.py)

Question. Almost every experiment behind this gene's annotations is synthetic-peptide
pharmacology — K17F, pE13F, [Pyr1]apelin-13, apelin-36 — much of it on rodent
receptors or in rodents. Two things follow only if the peptides are conserved: (a) that
those experiments are about the human gene product, and (b) that the PAINT node placed
at Tetrapoda is a safe home for the family's one IBD assertion.

Method. No multiple alignment is needed. Every bioactive apelin (apelin-36/-31/-28/
-17/-13) is released from the C-terminus of the precursor, so "the last n residues"
is an exact, alignment-free anchor; human apelin-13 is UniProt PEPTIDE 65..77 of a
77-residue precursor, i.e. exactly the last 13. All 333 UniProtKB entries that PANTHER
classifies into PTHR15953 were fetched, and per-column identity to human was measured over
the C-terminal 13 residues and, as an internal negative control, over an equally sized
window at the N-terminus (which falls inside the signal peptide, SIGNAL 1..22). The
script asserts that the human precursor is 77 aa before doing anything else.

Result.

Per column, the apelin-13 window splits cleanly in two:

So the invariant unit is not apelin-13 but the 12-mer R66–F77; position 65, the
glutamine that becomes the pyroglutamate of [Pyr1]apelin-13, is the one variable
position in the peptide (zebrafish carries Pro there).

Among the five Swiss-Prot members, all 77 aa:

Interpretation. Human, rat, mouse and bovine apelin-13 and apelin-17 are identical.
A "rat" or "bovine" apelin-17 in a pharmacology paper is, residue for residue, the human
gene product. Divergence is confined to apelin-36, i.e. to the part of the precursor that
is trimmed away. This is why the rodent-derived ISS rows and the peptide-pharmacology IDA
rows on this gene are on much firmer ground than a species caveat would suggest — and also
why they say nothing species-specific.

2. Does the human C-terminus carry the ACE2 substrate motif? (cterm_conservation.py)

Question. Apelin is inactivated by ACE2, which removes a single C-terminal residue.
Vickers et al. (PMID:11815627) derived a consensus from an alignment of ACE2 substrates:
"Pro-X((1-3 residues))-Pro-Hydrophobic, where hydrolysis occurs between proline and the
hydrophobic amino acid". Does human apelin match it, and does the family?

Result. The human C-terminal hexapeptide is KGPMPF; P74-M75-P76-F77 is
Pro-X(1)-Pro-Phe, so the motif is present and predicts removal of Phe77 only —
which is what ACE2 is reported to do. Across the family the motif is present in
328/333 (98.5 %) members, and the terminal residue is Phe in the same 328/333
(98.5 %); the five exceptions terminate in E (2), P, Y and S.

Interpretation. The single residue that ACE2 removes to inactivate apelin is the same
residue UniProt flags as a determinant of G(i)-versus-β-arrestin balance (SITE 77,
ECO:0000269|PubMed:38428423), and it is invariant across the family. The catalytically
relevant claim in this review is therefore a retention claim, not a loss claim: human
apelin retains the C-terminal Phe its rodent and bovine orthologs carry, which is what
makes their pharmacology transferable and what makes ACE2 a regulator of the human
peptide too.

3. Provenance of the inferred rows (resolve_withfrom.py, check_goa_reconciliation.py)

resolve_withfrom.py reads the review YAML — never a retyped list — and resolves every
supporting_entities identifier against a live database, writing
withfrom_resolved.tsv and the source_entities.yaml scaffold that the
propagation_review blocks are built from.

• 19 of the 43 rows in the finished review carry supporting_entities — 18 of the 41
  GOA-seeded rows, plus the one NEW ISS row this review adds.
• 17 of those require propagation_review: IEA ×10, ISS ×6, IBA ×1.
• The remaining 2 are IGI rows (GO:0010629, GO:0040037) whose supporting_entities
  are genetic-interaction partners, not propagation donors.
• 13 distinct identifiers, 0 unresolved.

Resolutions that mattered to the review:

The node resolution is worth spelling out, because it is what the IBA row rests on.
PTN001041490 is the only node-level annotation in the whole family: an IBD for
GO:0005576, not negated, placed at taxon:32523, which NCBI Taxonomy gives as Tetrapoda
(rank CLADE). Human is inside that clade, so the target inherits. Its three gene-level
seeds are rat Apln, bovine APLN and human APLN itself — the target appearing in its own
WITH/FROM, which is the expected marker that experimental grounding exists on the target,
not a circular transfer.

The two RNAcentral ids only resolve informatively with their _9606 suffix kept: the bare
URS accessions describe cross-species sequences ("ncRNA from 16 species") and hide which
human miRNA the row means. They are exactly the two miRNAs the cited paper is about.

check_goa_reconciliation.py re-derives the join key
(GO id, evidence code, reference, normalized WITH/FROM) from APLN-goa.tsv and from the
review YAML and requires a bijection. Current state: 41 GOA rows, 41 reviewed YAML
entries, reconciliation OK, plus whatever NEW rows the review adds, which the script
lists and excludes from the join.