ABRA (STARS) — curation notes

Human ABRA, UniProt Q8N0Z2, 381 aa, HGNC:30655. Also called STARS (striated muscle
activator of Rho signaling) and MS1 (myocyte stress 1).

1. What the protein does

The mechanism is worked out and consistent across three papers from one group plus
independent confirmation, but almost all of it is rodent.

2. What the human GO record actually contains

See ABRA-bioinformatics/RESULTS.md for the reproducible audit. Headline numbers: 16 rows,
2 experimental (one Y2H IPI, one LIFEdb IDA), 14 transferred; and of the transfers,
mouse Abra is the WITH/FROM source for 8 rows under four different identifier styles
(UniProtKB:Q8BUZ1, ensembl:ENSMUSP00000051973, MGI:MGI:2444891, InterPro:IPR026111),
rat Abra for 3 more. No human experiment has established any molecular function, biological
process or native localisation for this protein. The mechanism above is entirely mouse,
rat, zebrafish and C2C12; the human contribution to the literature is expression profiling.

The propagation topology is clean, though: PANTHER PTHR22739 has 4 reviewed members, all
named ABRA, all in subfamily SF20, so the IBA node PTN001100454 and the ISS transfers sit
inside a one-to-one ortholog group with no paralog to confuse them.

3. The plasma-membrane annotation

GO:0005886 IDA under GO_REF:0000054 (LIFEdb GFP-fusion survey) is the one annotation
that is not a rodent transfer, and it is the one I do not believe:

Marked MARK_AS_OVER_ANNOTATED rather than REMOVE: the observation itself may be real for
an overexpressed fusion, and GO_REF:0000054 is a pipeline reference with no full text to
read against.

4. The GO gap: the MRTF nuclear-import step is missing from human

The mechanistic heart of what ABRA does — driving MRTF-A/MRTF-B into the nucleus — has no
human GO annotation. Mouse has GO:0006606 protein import into nucleus IDA from
PMID:15798203, but that term says STARS is imported, which is not what the paper shows;
STARS causes MRTFs to be imported. The correctly-directed term is
GO:0042307 positive regulation of protein import into nucleus. Because the mouse
annotation used the wrong-direction term, the step was never propagated to human, and the
human record jumps straight from actin binding to positive regulation of transcription by RNA polymerase II with the causal middle missing. Added as a NEW annotation and flagged
for the mouse record too.

5. Caveats on the affinage record

self_evaluation_pairwise: win with clear trust gates, all 12 citations are real numeric PMIDs,
and the mechanistic narrative
is accurate. One over-attribution:

"Its actin-cytoskeleton-regulating activity is conserved through the C-terminal Costars
domain, whose function in actin organization and motility is preserved across species
PMID:20940261."

PMID:20940261 is about a different gene. The Costars protein is 82 residues
PMID:20940261 and the rescue used
PMID:20940261. The human counterpart of that 82-residue protein is ABRACL (Q9P1F3,
81 aa) — RESULTS.md Q4 confirms ABRACL and ABRA are the only two reviewed human proteins
carrying Pfam PF14705, and ABRACL is essentially the domain alone. Sharing a domain is not
sharing a phenotype; nothing from that paper is used to support an ABRA annotation here.

6. Things I deliberately did not annotate