Consistency: Deep research ↔ review YAML ↔ PN annotation consistent. EDEM2 = soluble ER-lumenal GH47 protein that catalyzes the first trimming step Man9→Man8B (E117-dependent, in an obligate disulfide complex with TXNDC11), committing substrates to gpERAD upstream of EDEM1/EDEM3; does NOT bind SEL1L. The negatedGO:0004571 IDA from PMID:15537790 (recombinant "no activity") is correctly retained as superseded by the positive GO:0004571 IMP (PMID:25092655) + E117Q mutant. Review and notes flag this explicitly. No contradictions.
PN story / NEW pressure: subtype→GO:1904382 marked more_specific_than_existing_goa. Note GOA already has GO:1904380 (ER mannose trimming, IMP+TAS+IEA) but NOT GO:1904382 (the ERAD-specific child) for EDEM2 — so the PN projection of GO:1904382 to EDEM2 is a defensible, more-specific real annotation (EDEM2 demonstrably trims in the gpERAD pathway, PMID:25092655). GO:1904382 verified real (OLS, non-obsolete). This is the one genuine ADD candidate among the EDEMs. [REF/MAP]
Mapping strategy: This gene does sharpen the node: EDEM2 is the catalytic first-step mannosidase, the clearest "Mannose trimming" exemplar. subtype→GO:1904382 is appropriate and adds specificity over existing GO:1904380. group→GO:0006487 protein N-linked glycosylation (more_specific_than_existing_goa) is broader/upstream than EDEM2's degradative trimming and is a loose fit — borderline over-reach for a degradation-arm enzyme.
Evidence alignment: Strong overlap on PMID:25092655 (shared with EDEM1/3). Review adds the mechanistic TXNDC11 disulfide paper (PMID:32065582), folding-state dependence (PMID:30374462), and pathway model (PMID:39654396) beyond the PN row.
Verdict: Consistent and well-curated; catalytic (first-step) EDEM, NOT-mannosidase handled correctly. subtype→GO:1904382 is a defensible more-specific ADD; group→GO:0006487 is a loose/broad fit.