IRE1 review notes

2026-09-02 Update: nuclear-localization annotation (GO:0005634, IDA, PMID:17035634)

Audited the existing review for oversights. The IDA annotation of GO:0005634 (nucleus)
from PMID:17035634 had been marked UNDECIDED with the stated reason "Unable to access
PMID:17035634 to verify the nuclear localization claim." This was incorrect: the
publication is cached in this repository (publications/PMID_17035634.md) and its
abstract directly and unambiguously supports the annotation.

Goffin et al. 2006 (Mol Biol Cell) show that Ire1p's cytoplasmic linker region contains
an 18-residue nuclear localization sequence (NLS) recognized by both importin alpha
(Kap60p) and multiple importin beta homologues, that this NLS drives Ran-GTPase-dependent
nuclear import of Ire1p (or an NLS-GFP reporter) in vivo, and that NLS-disrupting point
mutations impair ER-stress-induced HAC1 mRNA splicing:

PMID:17035634

PMID:17035634

PMID:17035634

Changed the annotation's action from UNDECIDED to KEEP_AS_NON_CORE: the nuclear
pool and NLS-dependent import are real and evidence-backed (so UNDECIDED, reserved for
cases where the evidence genuinely cannot be assessed, was not appropriate), but IRE1's
best-established, defining catalytic activities (kinase trans-autophosphorylation and
HAC1 pre-mRNA endoribonuclease splicing) are ER-membrane events, so nuclear import is
kept as a non-core regulatory/trafficking aspect rather than promoted into
core_functions.

All other existing annotations, core_functions, and the top-level description were
reviewed and found sound and well-supported; no other changes made.

2026-09-04 — SFT review consistency after the nuclear-localization re-review

The SFT prediction review still classified GO:0005634 (nucleus) as NPI and asserted
that IRE1 is never found in the nucleus after the main gene review retained the IDA
annotation from PMID:17035634 as KEEP_AS_NON_CORE. The benchmark policy treats an
exact term retained with a positive AIGR action as CNN (correct but not novel), even
when the function or location is non-core. The standard SFT audit identified this
single deterministic category conflict across its 95-gene cohort.

The categorical exclusion is also too strong for the cached abstract, which reports
NLS-dependent localization and mutational effects on signaling PMID:17035634. This consistency
repair follows the existing annotation review; it does not infer that nuclear
localization is IRE1's principal location or resolve the mechanism of trafficking of
the intact membrane protein. The reference cache remains abstract-only.