ISDS Skeletal-Dysplasia Classification
dismech tags genetic skeletal disorders with their group in the Nosology of Genetic Skeletal Disorders, the expert-consensus nosology maintained by the Nosology Committee of the International Skeletal Dysplasia Society (ISDS). The encoded edition is the 2023 revision — the eleventh — with 771 entries, 552 genes, and 41 groups:
Unger S, Ferreira CR, Mortier GR, et al. Nosology of genetic skeletal disorders: 2023 revision. Am J Med Genet A. 2023;191(5):1164-1209. PMID:36779427, doi:10.1002/ajmg.a.63132
It supersedes the 2019 revision (10th edition; 461 disorders, 437 genes, 42 groups; Mortier et al., PMID:31633310), which dismech encoded first and which is retained as provenance — see Revision handling.
| Artifact | Path |
|---|---|
| LinkML enum (41 groups + 4 deprecated) | src/dismech/schema/classifications/isds_skeletal_nosology.yaml |
| Assignment class | ISDSNosologyAssignment in src/dismech/schema/dismech.yaml |
| Slot | classifications.isds_skeletal_category |
| Curator guidance | .claude/skills/disease-classification/SKILL.md |
What the nosology is
The ISDS Nosology is the reference list of recognized genetic skeletal entities. A disorder is included only if it has significant skeletal involvement, is published and catalogued (PubMed/OMIM), has a proven genetic basis, and has nosologic autonomy — it is an independent entity rather than a variant of an existing one. That last criterion is why the nosology is worth encoding: it is a curated statement about which skeletal entities are real and distinct, not merely a convenient bucketing.
Groups mix organizing principles deliberately:
| Groups (2023) | Organizing principle |
|---|---|
| 1-8 | Shared causal gene, gene family, or biosynthetic pathway (FGFR3, type 2 / type 11 collagen, sulfation, joint dislocations, filamins, proteoglycan core proteins, TRPV4) |
| 9-14 | Radiographic — which segment of the growing bone is affected (epiphyseal, metaphyseal, spondylo-), plus the skeletal ciliopathies at 10 |
| 15-19 | Limb-segment pattern (mesomelic, acromesomelic, acromelic) and the brachydactylies |
| 20-23 | Bent bones, primordial dwarfism, lysosomal storage, chondrodysplasia punctata |
| 24-28 | Bone density and mineralization (osteopetrosis, osteosclerosis, OI/fragility, mineralization, PTH signaling) |
| 29-32 | Osteolysis, disorganized development, overgrowth, inflammatory |
| 33-37 | Dysostoses (cleidocranial, craniosynostosis, craniofacial, vertebral/costal, patellar) |
| 38-41 | Limb reduction, split hand/foot, polydactyly-syndactyly, synostoses |
Entry numbers (NOS <group>-<entry>)
Each row of the 2023 Table 1 carries its own identifier in the form
NOS 05-0060 / NOS 17-0150 — the two-digit group number, then a
four-digit within-group sequence. It is the nosology's own row identifier,
not something this repo invents, and it is the most precise way to point at
a single listed entity: group numbers move between revisions and disorder
names get rewritten (see Revision handling), but the
row number pins which line of which table a notes: claim came from.
The group prefix is revision-specific, exactly like a bare group number:
NOS 17-0150 only resolves once you know it is a 2023 row, since the same
entity carried a 2019 group-38 number. So a quoted NOS id needs its revision
stated alongside it, just as the group name does.
Quoting it is optional and currently rare — most notes: cite the group name
plus the listed disorder name, which is enough. Use it when the entity is
easy to confuse with a neighbour, as Pseudohypoparathyroidism does
(NOS 17-0150, the row whose name changed between revisions). Always pair it
with the OMIM number when the source row has one: OMIM survives renames, and
the two together make the identification checkable without the paper.
Revision handling
The nosology is revised every few years (2002, 2007, 2011, 2015, 2019, 2023). dismech tracks it with three LinkML constructs so a revision bump never invalidates existing annotations:
Keys are stable identifiers. A permissible-value key is not renamed or
renumbered when a revision renames or renumbers its group — the key identifies
the group across editions, while the revision's own number and name live in the
description. Group numbers are emphatically not stable: the brachydactyly
groups moved from 37/38 (2019) to 18/19 (2023), and the lysosomal storage group
from 27 to 22. That is exactly why numbers are not part of any key.
Superseded names become structured synonyms. LinkML permissible values
accept structured_aliases, so the old name is retained with its predicate and
a source citing the revision that used it — a search for the old name still
resolves, and the citation says which edition it came from:
craniosynostosis_syndromes:
structured_aliases:
- literal_form: Craniosynostosis syndromes
predicate: EXACT_SYNONYM
source: PMID:31633310
description: >-
Name of this group in the 2019 revision (10th edition), where it was
group 33.
description: >-
Group 34 (2023 revision): Syndromes featuring craniosynostosis. ...
34 of the 41 active groups carry such an alias. The seven that do not split into two distinct cases:
- Four kept both their name and their number across the two revisions, so there is nothing to alias: metaphyseal dysplasias, spondylometaphyseal dysplasias, severe spondylodysplastic dysplasias, and acromesomelic dysplasias.
- Three are genuinely new in 2023 and have no 2019 name to record: proteoglycan core protein disorders, osteosclerotic disorders, and skeletal disorders of parathyroid hormone signaling cascade. (The first two are the merge targets of the four deprecated groups; the third is new outright.)
The four deprecated 2019 values also carry aliases, which is why a naive count over all 45 permissible values returns 38 rather than 34.
Dissolved groups are deprecated, not deleted. When a revision merges a group
away rather than renaming it, the old value is kept with deprecated: and
deprecated_element_has_possible_replacement pointing at its successor —
possible rather than exact because a merge makes the successor broader than
the value it replaces:
perlecan:
deprecated: >-
Group 5 "Perlecan group" of the 2019 revision (PMID:31633310). The 2023
revision (PMID:36779427) dissolved it into the new group 7 "Proteoglycan
core proteins disorders", which also absorbed the former Aggrecan group.
Retained so existing assignments remain resolvable; do not use for new
curation.
deprecated_element_has_possible_replacement: proteoglycan_core_protein_disorders
Four 2019 groups are deprecated on that basis: Perlecan and Aggrecan (merged into Proteoglycan core proteins disorders), and Neonatal osteosclerotic dysplasias and Other sclerosing bone disorders (fused into Osteosclerotic disorders). One group is new in 2023 with no 2019 counterpart: Skeletal disorders of parathyroid hormone signaling cascade.
Dyadic naming
The headline change in 2023 is dyadic naming: each phenotypic entity is systematically paired with its causal gene — "Geleophysic dysplasia, ADAMTSL2-related" — replacing list numberings and eponyms, which the committee considers "more informative and less prone to errors."
That is a claim about disease-entity naming and identity, not about this
classification axis, so dyadic names are not adopted as dismech entry
name: values. They are instead recorded as synonyms, which captures the
naming without committing dismech to it — whether dismech should adopt dyadic
naming as primary remains a scope/identity question for the
design-decision register.
50 entries carry a dyadic synonym, added under a deliberately strict 1:1 rule:
- the entry's name stem must match a 2023 row stem (dyadic suffix stripped),
- the row's gene must be one of the entry's curated genes, and
- exactly one 2023 row may match that stem.
The third condition does the real work. Where 2023 splits an entity into several gene-specific rows, the dyadic names denote subtypes, not synonyms of the whole — Coffin-Siris (5 rows), Cornelia de Lange (5), Meckel (6), Loeys-Dietz (6), Adams-Oliver (6), geleophysic dysplasia (3), sclerosteosis (SOST and LRP4). Attaching any single one of those to the umbrella entry would assert a false identity, so those 19 entries are skipped.
Two cases are worth knowing about:
- Multiple sulfatase deficiency is excluded because the published table
reads "Multiple sulfatase deficiency, SUMF-related" while its own gene column
says
SUMF1. Recording a misspelt gene symbol as a searchable synonym would do more harm than the synonym is worth. - Spondylodysplastic Ehlers-Danlos syndrome is excluded because the dismech entry is a three-gene umbrella (B3GALT6, B4GALT7, SLC39A13) that the 2023 revision splits across three rows in two different groups (5 and 13). That also means its current group assignment needs re-checking in the verification pass.
Known gap
The exemplar disorders named in each group description were transcribed from Table 1 of the 2019 revision. They have been renumbered for 2023 and corrected where the 2023 paper explicitly says a disorder moved (e.g. trichorhinophalangeal dysplasia types 1/3 and Langer-Giedion, which left the acromelic group for Brachydactylies as part of syndromes). A full re-transcription against the 2023 table — 774 rows, already extracted — is outstanding, as is re-verifying the existing per-entry assignments against it.
How to assign it
classifications:
harrisons_chapter:
- classification_value: GENETICS_ENVIRONMENT_DISEASE
isds_skeletal_category:
- classification_value: fgfr3_chondrodysplasia
notes: >-
ISDS Nosology of Genetic Skeletal Disorders, 2023 revision
(Unger et al., PMID:36779427), group 1 "FGFR3 chondrodysplasias";
listed as "Achondroplasia, FGFR3-related".
Three rules follow from how the nosology is built:
- One group per disorder. The committee lists each disorder exactly once, "to avoid redundancy in the Nosology". Overlap between groups is handled by Table 1's see also cross-references, not by dual membership. The slot is multivalued only to accommodate a dismech entry that lumps several distinct nosology disorders.
- Assign only to listed disorders. The enum is a transcription of an expert nosology, not an inference engine. Many disorders with skeletal phenotypes were deliberately excluded (chiefly for lack of significant skeletal involvement); do not extend the classification to them. An unambiguous subtype or synonym of a listed disorder is fine.
- Use
notes:, notevidence:. The paper's PubMed record is abstract-only. No exact-quote snippet from the abstract can support a per-disorder group placement — the abstract states only that the nosology exists and contains 771 entries across 552 genes in 41 groups. Quoting it for a specific assignment would be a snippet that does not support its claim. This mirrors the ICIMD convention.
Cross-group traps
The gene most associated with a disorder is often not what decides its group. Frequently mis-assigned cases:
All numbers below are 2023 group numbers. (Their 2019 equivalents differ — see Revision handling — which is why the guidance is to cite a group by name and revision, never by a bare number.)
| Disorder | Correct group (2023) | Tempting but wrong |
|---|---|---|
| Crouzon with acanthosis nigricans, Muenke craniosynostosis | 34 syndromes featuring craniosynostosis | 1 FGFR3 chondrodysplasias |
| LADD syndrome | 40 polydactyly-syndactyly-triphalangism | 1 FGFR3 chondrodysplasias |
| Hajdu-Cheney syndrome | 29 osteolysis | 26 OI and bone fragility |
| Brachydactyly-hypertension (Bilginturan) | 19 brachydactylies as part of syndromes | 17 acromelic dysplasias |
| Weyers acrofacial (acrodental) dysostosis | 35 craniofacial dysostoses | 10 skeletal ciliopathies (with Ellis-van Creveld) |
| Spondylodysplastic Ehlers-Danlos syndrome | types 1 (B4GALT7) and 2 (B3GALT6) → 5 multiple joint dislocations; type 3 (SLC39A13) → 13 SE(M)D | one group for all three |
| Acrodysostosis (PRKAR1A, PDE4D) | 17 acromelic dysplasias | 28 PTH signaling — see below |
| Albright hereditary osteodystrophy (GNAS) | 17 acromelic dysplasias | 28 PTH signaling; or the brachydactyly group, where the 2019 revision listed it as "Pseudohypoparathyroidism type IA" (2019 group 38, whose 2023 successor is group 19) |
| Meier-Gorlin syndrome (ORC1, ORC4, ORC6, CDT1, CDC6, GMNN, CDC45, MCM3/5/7, GINS2) | 21 primordial dwarfism and slender bones | 37 patellar dysostoses — the disorder's own name begins "ear-patella" |
The Meier-Gorlin row is the strongest name-based trap in the set. The 2019 revision called it ear-patella-short stature syndrome, so the patellar dysostoses group looks obviously right — but all eleven of its rows are in group 21, and the 2023 rename to "ear-patella-primordial short stature syndrome" is the tell. Group 37 holds only ischiopatellar dysplasia (TBX4), nail-patella syndrome (LMX1B), and genitopatellar syndrome (KAT6B). Both the group 21 and group 37 enum descriptions state the placement explicitly, in both directions.
The GNAS row is a rename and a move, which makes it easy to miss: OMIM 103580
appears in the 2019 revision as Pseudohypoparathyroidism type IA in group 38
(brachydactylies with extraskeletal manifestations) and in the 2023 revision as
Albright hereditary osteodystrophy, GNAS-related in group 17. Searching the
2023 table for "pseudohypoparathyroidism" returns nothing — the entity did not
disappear, it was renamed. The nosology's only other GNAS entries are polyostotic
fibrous dysplasia / McCune-Albright and progressive osseous heteroplasia, both in
group 30. No GNAS disorder is in group 28.
For the two dismech entries this affects:
Pseudohypoparathyroidismis assigned to group 17, scoped by itsnotes:to the PHP1A subtype — that subtype is the listed entity (OMIM 103580). PHP1B and PHP2 are not in the nosology at all, so the dismech entry is broader than its counterpart; the note says so rather than implying the whole umbrella was transcribed.Pseudopseudohypoparathyroidismis deliberately left unassigned. Its OMIM (612463) has no row in the 2023 table. Do not read that omission as a claim that PPHP falls outside the nosology's scope — the listed row's label is Albright hereditary osteodystrophy, which PPHP manifests (AHO without hormone resistance), so label and OMIM disagree about coverage. Assigning on the label would be inference; the honest state is unassigned. If a future revision lists PPHP separately, group 17 is where it would go. Note that MONDO disagrees:MONDO:0012912pseudopseudohypoparathyroidism is a descendant ofMONDO:0019695acromelic dysplasia, so the ontology already places it in this family. That is not evidence the entry was simply overlooked here — it is a second classification making its own call, and this axis transcribes the ISDS committee rather than deferring to MONDO (which is also why the group carries nomeaning:; see MONDO mappings).
The Acrodysostosis row is a live trap in the other direction. Acrodysostosis is
mechanistically a PTH/PTHrP-signalling disorder, so group 28 looks right — but
the 2023 table lists both Acrodysostosis, PRKAR1A-related and Acrodysostosis,
PDE4D-related under group 17, acromelic dysplasias. Group 28 contains the
PTHR1/SIK3/PTHLH disorders (Jansen and Csukasi-Krakow metaphyseal dysplasia,
Blomstrand dysplasia, Eiken dysplasia, PTHLH brachydactyly and osteolysis).
Reasoning from mechanism to placement is exactly the inference this axis must
not make: the nosology is a transcription of expert placement, not a derivation
from pathway.
Relationship to other classification axes
isds_skeletal_category is orthogonal to the other axes and should be set
alongside them, not instead of them:
harrisons_chapter— usuallyGENETICS_ENVIRONMENT_DISEASE.icimd_category— the storage disorders in ISDS group 22 (lysosomal storage diseases with skeletal involvement) are inborn errors of metabolism and also carry an ICIMD group; likewiselysosomal_storage_category.mechanistic_category— ISDS group 10 members are ciliopathies, which is a mechanism statement rather than a nosology placement.
MONDO mappings
No group carries a meaning:. A meaning: would assert that the value is
an ontology class, which is never quite true: the ISDS group is a curated list,
the MONDO class is a defined extension, and the two rarely coincide. Where a
MONDO class denotes the same disease family, it is recorded as
close_mappings: instead.
Three of the 41 groups are mapped. The bar is deliberately high — a candidate is rejected if the MONDO class contains any entity that ISDS itself lists in a different group, since such a mapping would silently contradict the committee's own placement:
| Group | MONDO | Why it survives |
|---|---|---|
| 1 FGFR3 chondrodysplasias | MONDO:0019685 FGFR3-related chondrodysplasia |
Contains the group's members and, crucially, excludes the FGFR3 craniosynostoses (Muenke, Crouzon with acanthosis nigricans) that ISDS puts in the craniosynostosis group (34) |
| 8 TRPV4 disorders | MONDO:0018240 TRPV4-related bone disorder |
All descendants are group-8 members; parastremmatic dwarfism appears in the group's own SEMD Maroteaux row |
| 16 Acromesomelic dysplasias | MONDO:0019696 acromesomelic dysplasia |
All descendants are either group-16 members or entities ISDS does not list at all |
Rejected candidates, recorded so nobody re-proposes them:
| Group | Rejected MONDO | Straddle |
|---|---|---|
| 2 | MONDO:0022800 type 2 collagenopathy |
contains SMD 'corner fracture' type, listed in the SMD group (12) |
| 17 | MONDO:0019695 acromelic dysplasia |
contains the trichorhinophalangeal syndromes and Langer-Giedion (moved to group 19 in 2023), terminal osseous dysplasia (group 6), and short-rib thoracic dysplasia 9 (group 10). The earlier rationale also cited pseudohypoparathyroidism type 1A — that no longer applies, because 2023 renamed it Albright hereditary osteodystrophy and placed it inside group 17 |
| 24 | MONDO:0017198 osteopetrosis |
contains melorheostosis and osteopathia striata with cranial sclerosis, now in the osteosclerotic group (25) |
| 34 | MONDO:0015338 syndromic craniosynostosis |
contains cranioectodermal dysplasia, a skeletal ciliopathy (group 10) |
Several groups have no usable class at all: either the nearest MONDO term is obsolete (chondrodysplasia punctata) or it is narrower than the ISDS group, which mixes in entities falling outside it — the OI and bone fragility group holds cutis laxa and Singleton-Merten alongside osteogenesis imperfecta.
Note the asymmetry that makes close_mappings the right relation: in every
mapped case MONDO is broader, because ISDS lists only entities meeting its
inclusion criteria. A mapping is rejected not for breadth but for crossing
into another group.
Entities the nosology lists but declines to decompose
Fanconi anemia is the worked precedent. It is listed in the limb hypoplasia - reduction defects group (39 in the 2019 revision, 38 in 2023), but it is the only Table 1 row whose name carries a "(see note below)" pointer and the only one whose gene column reads "Several" rather than naming loci. The group-39 group footnote says the complex genetic basis of Fanconi anemia and its complementation groups "is acknowledged but not further listed in this nosology", referring readers to OMIM or specialized reviews.
Read that precisely: the caveat is about genetic decomposition, not membership. The committee does not question whether Fanconi anemia belongs in the nosology — it lists it without qualification on that point, and its skeletal phenotype (radial ray and thumb reduction defects) is exactly what that group is for. What it declines to do is enumerate FA-A, FA-C, FA-D2 and the rest.
The curation rule that follows: assign the group to the entity, record the
committee's caveat in notes:, and do not derive per-complementation-group
placements from Table 1. Apply the same treatment to any future row the
committee flags this way.
Refreshing to a later revision
The ISDS Nosology is revised every few years (2002, 2007, 2011, 2015, 2019,
2023). The enum is deliberately pinned to a single named revision, with the
revision recorded in the schema description and in every assignment's notes:.
When moving to a later edition, add or rename permissible values rather than
silently redefining existing ones, and update the notes: template so an
assignment always says which revision it came from — group numbering is not
guaranteed stable across revisions.