The GO:0051082 prediction remains CNN for the biological concept of binding non-native glycoprotein substrates. Recombinant rat UGGT retains the substrate preference of the liver enzyme [PMID:10764828, “The purified recombinant enzyme shows the same preference for unfolded polypeptides”]. The exact term has an IDA GOA row, and the current main review explicitly affirms the underlying recognition/binding while marking a standalone unfolded-protein-binding annotation over-annotated. Substrate binding is not disproved by preferring the glucosyltransferase activity as the curated molecular function, nor does the predicted binding term assert a separate folding-chaperone mechanism.
The benchmark policy separates ontology status from biological correctness. Its frozen ontology already treats release-specific identifiers separately from the underlying biological concept. An explicit supported-wrapper adjudication now preserves CNN for this gene/term when the reference action is MARK_AS_OVER_ANNOTATED; other rejection actions or accepted negations trigger renewed review. The main review, raw model identifier/label, and GOA snapshot are unchanged. Cached PMID:10764828 is abstract-only; the claim here is anchored to the explicit rat enzyme/substrate findings in that abstract and its existing IDA annotation, without inventing uninspected assay details.
LSP was considered but not used: the preferred glucosyltransferase activity is a different catalytic concept, rather than a refinement of the predicted binding specificity. The distinction concerns the usefulness of an annotation, not refutation of the experimentally supported binding concept.
GOA changes since the original review:
- New rows (4): GO:0003980 EXP PMID:1533626; GO:0005788 ER lumen ISO and GO:0005793 ERGIC
ISO (both from human UGGT1 Q9NYU2); GO:0006487 protein N-linked glycosylation IBA
(PTN000132051).
- Retired rows (6): GO:0051082 IBA/ISO/IDA (QuickGO confirms GO:0051082 is obsolete),
GO:0016740 and GO:0016757 (keyword IEA), GO:1901135 (ARBA IEA). Reviews kept; a
retirement sentence added to each. The three GO:0051082 rows keep MARK_AS_OVER_ANNOTATED
(recognition of non-native clients is the substrate-selection step of the
glucosyltransferase, not a separate holdase/chaperone activity), so no
proposed_new_terms entry is needed: the core MF has a GO term (GO:0003980).
PENDING resolutions:
- GO:0003980 EXP PMID:1533626 -> ACCEPT PMID:1533626
- ER lumen ISO and ERGIC ISO -> ACCEPT, consistent with rat IDA rows PMID:11535823.
- GO:0006487 IBA -> KEEP_AS_NON_CORE: UGGT acts on asparagine-linked glycans (N-glycan
processing is part_of protein N-linked glycosylation), but the transient
reglucosylation serves quality control rather than glycan biosynthesis.
Action changes:
- GO:0005515 IPI PMID:11278576 (with SELENOF Q923V8): MARK_AS_OVER_ANNOTATED -> REMOVE.
The complex is real PMID:11278576, but no specific MF follows;
complex membership is captured by the GO:0032991 row.
- GO:0005515 IPI PMID:10764828 (with Q923V8): MARK_AS_OVER_ANNOTATED -> REMOVE (same
reasoning; the cached abstract does not mention SELENOF).
- GO:0051084 'de novo' post-translational protein folding TAS: KEEP_AS_NON_CORE -> MODIFY
to GO:0034975 protein folding in endoplasmic reticulum, matching the protein folding TAS row.
- 17 rows carrying boilerplate "Manual review ... Retained as supported or plausible"
text were rewritten with specific summaries and verbatim supporting quotes; actions
unchanged. Protein secretion ISO stays KEEP_AS_NON_CORE: QuickGO shows the mouse
donor row (Q6P5E4) is IMP from PMID:20498017 PMID:20498017.
Open questions:
- Whether GO should have an activity term for glycoprotein folding-sensor recognition;
for now it is treated as intrinsic to GO:0003980.
Review follow-up 2026-10-19 (PR #4624): the GOA refresh retired the IDA GO:0051082 unfolded protein
binding row (PMID:10764828) for Uggt1, so the manual BioReason-Pro SFT adjudication rationale
"GOA carries the exact term with IDA" was no longer true. The CNN assessment stands; the rationale
in Uggt1-sft-predictions.yaml and its frozen copy in scripts/auto_review_sft_predictions.py
(MANUAL_OVERRIDES) now say GOA carried the term until the 2026-10 refresh retired the row, a
curation-status change that does not refute the binding.