Disorders of Protein and Lipid Glycosylation (ICIMD category 18)

Grouped on the shared mechanism family — a lesion in glycoconjugate metabolism, with hypoglycosylation (or, for the deglycosylation arm, failure to clear a released glycan) as the common intermediate lesion — and on an established nosological convention: category 18 of the International Classification of Inherited Metabolic Disorders (ICIMD; Ferreira et al. 2021, PMID:33340416), whose own subdivision into N-linked, O-linked, lipid, and multiple-pathway glycosylation disorders is what the nested sub-groupings mirror. This grouping points down to nested sub-groupings rather than listing member diseases flatly, following the pattern of Inborn_Errors_of_Immunity over the IUIS tables. It adds no direct disease members of its own. Each child keeps its own, narrower NECESSARY criterion; this parent's criterion is the disjunction of theirs, so any disease that satisfies its own sub-grouping necessarily satisfies this one. Why this exists rather than being folded into the existing Congenital_Disorders_of_Glycosylation grouping. That grouping is deliberately scoped to protein N-glycosylation and its precursor and trafficking supply lines — a boundary its own rationale states and defends — while binding MONDO:0015286 (congenital disorder of glycosylation), which is the broader concept covering O-linked, GPI-anchor, and glycosphingolipid defects as well. A grouping cannot both be deliberately narrower than a concept and exactMatch it. Creating the umbrella resolves that: the exactMatch to MONDO:0015286 moves here, and the N-glycan grouping is downgraded to skos:narrowMatch, which still counts as curated for MONDO coverage. Naming. ICIMD titles category 18 "Congenital disorders of glycosylation", but that name is already taken in kb/groupings/ by the N-glycan-scoped child and grouping names are foreign-key targets, so renaming it would break existing references and history records. This grouping therefore takes a descriptive name; "ICIMD category 18" is carried in display_name so the correspondence is not lost. Boundary — two exclusions worth recording, because both are things a reader would expect to find here. Glycogen storage diseases are out: glycogen is a storage homopolysaccharide, not a glycoconjugate, and ICIMD places disorders of glycogen metabolism outside the glycan-metabolism categories explicitly. The mucopolysaccharidoses are also out: they are lysosomal catabolism of glycosaminoglycans, already held by Mucopolysaccharidoses beneath Lysosomal_Storage_Disorders, and ICIMD places glycosaminoglycan *synthesis* under O-linked protein glycosylation rather than placing GAG degradation in category 18. Both are glycobiology in the loose sense; neither is a defect of glycan attachment to a protein or lipid.

MONDO:0015286 is the broad congenital-disorder-of-glycosylation class, spanning N-linked, O-linked, GPI-anchor, and glycosphingolipid defects. That is this grouping's concept, not the narrower N-glycan concept of the nested Congenital_Disorders_of_Glycosylation child, which now carries skos:narrowMatch against the same term. MONDO descendants without DisMech entries are curation gaps rather than a reason to downgrade the predicate.

MONDO consistency: inconsistent Re-derived 2026-09-18 over all 38 transitively held members: 24 are hierarchical descendants of MONDO:0015286 and 14 are not. Recorded as INCONSISTENT because that is what was measured; the mapping predicate is deliberately left at skos:exactMatch pending a decision, and this note is the evidence for that decision rather than a pre-emption of it. Method, reproducible without a local MONDO build: each member's own MONDO term is resolved to its ancestor set via the OLS hierarchicalAncestors endpoint and MONDO:0015286 looked up in it — one bounded upward walk per member (13-38 terms) instead of one unbounded downward expansion of the class. The OAK ols:mondo adapter raises NotImplementedError for both directions, but the REST endpoint underneath answers ancestors. The 14 divide into two kinds, and only one is a defect. (1) Genuine scope difference: MONDO scopes congenital disorder of glycosylation narrower than ICIMD category 18 does. The two deglycosylation members sit under MONDO:0031376 congenital disorder of deglycosylation, a *sibling* of MONDO:0015286 beneath MONDO:0019052 inborn errors of metabolism, and five of the six GPI-anchor members are likewise outside the class. ICIMD puts both inside category 18; MONDO does not. (2) Classification by clinical presentation rather than mechanism: Dystroglycanopathy sits under muscular dystrophy, Autosomal Recessive Cutis Laxa Type 2A under cutis laxa and hereditary skin disorder (verified against its direct parents, which carry no CDG term), the SLC10A7 and EXTL3 entries under skeletal system disorder, the PIG-gene entries under the epilepsies, and VPS51-CDG under pontocerebellar hypoplasia. These are candidate MONDO term requests, not membership errors. What this means for the predicate: the umbrella's concept properly *contains* the MONDO class rather than equalling it, which is skos:broadMatch and not skos:exactMatch. The reason it has not simply been flipped is that the change is not free — per CLAUDE.md a broadMatch does not count as curated for MONDO coverage, and scripts/grouping_mondo_gaps.py derives descendant gaps only from an exactMatch, so flipping it here after the N-glycan child was downgraded to narrowMatch would leave MONDO:0015286 with no exactMatch anywhere in the family and silently end gap derivation for the whole CDG branch. Semantics and tooling disagree; that is a design call for the decision register, not a predicate to change twice in two days.