Skeletal dysplasia claim–evidence review (2026-07-25)

Correctness review of ten skeletal dysplasia entries, focused on whether each curated claim is actually supported by the evidence attached to it.

Scope and method

Entries reviewed (chosen to span the mechanistic classes: FGFR3 gain of function, type II/X collagenopathy, sulfate transport, osteoclast protease, mineralization, transcription factor):

Entry	Gene	Mechanism class
Achondroplasia	FGFR3	RTK gain of function
Thanatophoric Dysplasia Type 1	FGFR3	RTK gain of function (severe allele)
Kniest Dysplasia	COL2A1	Type II collagenopathy / ER stress
Spondyloepiphyseal Dysplasia Congenita	COL2A1	Type II collagenopathy / ER stress
Metaphyseal Chondrodysplasia, Schmid Type	COL10A1	Type X collagenopathy / UPR
Diastrophic Dysplasia	SLC26A2	Sulfate transport / PG undersulfation
Campomelic Dysplasia	SOX9	Transcription factor haploinsufficiency
Cleidocranial Dysplasia	RUNX2	Transcription factor haploinsufficiency
Pycnodysostosis	CTSK	Osteoclast protease deficiency
Hypophosphatasia	ALPL	Mineralization / enzyme deficiency

Checks run:

linkml-term-validator validate-data … --labels over all ten — all 10 pass; every HP/GO/CL/CHEBI/MAXO/NCIT/HGNC id and label is correct.
linkml-reference-validator validate data … over all ten — all 10 pass; every snippet: is an exact substring of its cached reference.
Reference-cache existence check over all ~800 evidence items — no missing cache files, so no fabricated PMID/DOI/ORPHA/CGGV identifiers.
Manual claim-vs-evidence reading of every pathophysiology node, causal edge, prevalence record, inheritance block, and quantitative assertion.

Because (1)–(3) are clean, every finding below is semantic: the quote is real and the term is real, but the claim attached to it goes beyond what the quote says. These are exactly the failures the mechanical validators cannot catch.

Findings

1. Unsourced iPSC/UPR claim propagated across two entries — highest priority

Both Kniest_Dysplasia.yaml:184 and Spondyloepiphyseal_Dysplasia_Congenita.yaml:169 assert:

"recent iPSC-derived human cartilage models show that some COL2A1 mutations cause ER procollagen storage without engaging canonical UPR" (Kniest)

"human iPSC-derived cartilage models suggest that ER retention can occur without robust canonical UPR activation, consistent with an ER procollagen storage disorder rather than a classical UPR-driven apoptosis pathway" (SEDC)

No reference cited in either file mentions iPSC-derived cartilage. (Verified by grepping every cached reference in both entries for iPSC/induced pluripotent: zero hits.) The claim materially qualifies the entries' central mechanism — it is the sole basis for hedging the ER-stress/UPR-apoptosis model — so it should not stand unsourced. Either attach the primary source or remove the hedge.

2. Explanations over-claiming beyond truncated snippets (Hypophosphatasia)

Four evidence items quote a fragment and then assert something the fragment does not say. All four pass substring validation.

Location	Snippet	Explanation claims	Problem
`Failure to thrive in infancy`	`"CLINICAL CHARACTERISTICS: Hypophosphatasia is characterized by defective"`	"describes infantile HPP as presenting with rickets-like features and metabolic complications that lead to failure to thrive"	The quoted text (and the sentence it truncates) says none of this.
`Hypotonia`	`"children, clinical features include skeletal, respiratory and neurologic"`	"Review identifies neurologic features including hypotonia"	Source says "neurologic complications"; hypotonia is never named.
`Narrow chest`	`"most severe perinatal and infantile forms, results in 50-100% mortality,"`	"thoracic hypomineralization causing respiratory compromise, consistent with narrow chest as a key morphological driver"	Source sentence is about mortality; it says nothing about the thorax.
`Bowing of the long bones`	`"of TNAP and potent inhibitor of mineralization. Thus, HPP features rickets or"`	"Rickets and osteomalacia … underlie the bowing deformities of long bones"	Inference, not evidence; the quote never mentions bowing.

A fifth is a near-miss worth fixing while in the file: the Hypercalcemia snippet "vitamin B6-dependent seizures, hypercalcemia with" truncates one word short of "hypercalcemia with high morbidity, and mortality" — the explanation's "high morbidity" claim is in the source but was cut out of the quote.

3. Prevalence record internally contradicts its own class (Hypophosphatasia)

Hypophosphatasia.yaml:65-70, adult forms:

prevalence_class: ABOVE_1_IN_1000
rate_low: 32.258065        # = 1 in 3,100
rate_high: 196.850394      # = 1 in 508

ABOVE_1_IN_1000 means >100 per 100,000, but rate_low is 32.3 per 100,000. The source range straddles the class boundary, so a single class cannot be correct for it. The arithmetic is right; the class assignment is not.

4. Mechanism generalized from a different disease entity (Campomelic Dysplasia)

The pathophysiology node Reduced SOX9 Stability and Matrix Defects (Campomelic_Dysplasia.yaml:119-157) rests entirely on PMID:39854231, and asserts five downstream CD-specific radiographic features (bell-shaped thorax, 11 rib pairs, hypoplastic scapulae, cervical spine instability, scoliosis).

That paper studied a congenital-vertebral-malformation cohort and explicitly reports a TAM-domain variant "associated with mild skeletal dysplasia and scoliosis" — it positions these patients as a distinct, milder point on the SOX9 spectrum, not as campomelic dysplasia. The protein-stability mechanism may well generalize, but no CD-specific evidence is offered, and the node reads as though it were established in CD.

Also in this entry: the first evidence item of SOX9-Mediated Chondrogenesis Disruption (line 95-104) quotes a generic background sentence ("SOX9 is a crucial transcriptional regulator of cartilage development and homeostasis") and tags it HUMAN_CLINICAL; it is background prose, not clinical data.

5. A superseded mechanism presented as current (Schmid MCDS)

Metaphyseal_Chondrodysplasia_Schmid_Type.yaml:310 carries a Functional Haploinsufficiency node stating haploinsufficiency "is considered a proximate cause of the clinical phenotype", cited to a 2005 review (PMID:15880705).

The same entry's own evidence undercuts it: PMID:17403716's snippet reads "Unlike Col10a1 null mutants, transgenic mice (FCdel) … displayed typical characteristics of MCDS" — i.e. haploinsufficient/null animals do not phenocopy MCDS, which is the classic argument against the haploinsufficiency model. The rest of the entry (ER stress sufficiency, PERK/ATF4/Sox9, carbamazepine and PERK-inhibitor rescue) correctly models gain of function.

This is a case the schema already handles: the two competing models should be mechanistic_hypotheses groups (as Achondroplasia.yaml does for FGFR3 constitutive vs. ligand-dependent activation), not two co-equal nodes.

6. Frequency evidence attached to causal-link claims (Cleidocranial Dysplasia)

The RUNX2 Haploinsufficiency node carries eight downstream edges (lines 118-214), each with causal_link_type: UNKNOWN and evidence consisting solely of an Orphanet frequency row, e.g.:

- target: Carious Teeth
  causal_link_type: UNKNOWN
  evidence:
  - reference: ORPHA:1452
    snippet: "HP:0000670 | Carious teeth | Very frequent (99-80%)"
    explanation: Orphanet records carious teeth as very frequent in CCD.

A frequency row supports the phenotype and its frequency band; it says nothing about a causal edge from RUNX2 haploinsufficiency. All eight are duplicated verbatim in the phenotypes: block, where the same evidence is entirely appropriate. The edges should either get mechanistic evidence or be dropped.

Also: penetrance: COMPLETE (line 46) has no supporting evidence — the cited snippet establishes autosomal dominant inheritance only. (The same unsupported-penetrance pattern appears at Thanatophoric_Dysplasia_Type_1.yaml:35, where the cited snippet describes mutation counts and does not address penetrance or heterozygosity.)

7. Non-endochondral phenotypes parented to endochondral-ossification nodes

Diastrophic_Dysplasia.yaml:~316-355: Impaired Endochondral Ossification fans out to ~21 targets including Cleft Palate, Bifid Uvula, Absent Uvula, and Tracheobronchomalacia. Palatal-shelf fusion failure and airway (non-endochondral) cartilage softening derive from the general proteoglycan-undersulfation matrix defect, not from failed endochondral bone formation. Nineteen of these edges are INDIRECT_UNKNOWN_INTERMEDIATES with no edge-level evidence.
Spondyloepiphyseal_Dysplasia_Congenita.yaml:~262-275: Impaired Growth Plate Organization → Cleft Palate and Flat Face, same issue.

Re-parenting these to the upstream matrix-defect node would be more accurate and costs nothing structurally.

8. Quantitative claims in `notes`/`description` with no supporting quote

Entry	Claim	Status
Schmid MCDS `:54`	"estimated incidence of approximately 3-6 per million"	Not present in any of the entry's cited references. The two evidence items support only "rare" and "most common subtype". The record is `prevalence_class: UNKNOWN`, so the figure also contradicts its own structured field.
Schmid MCDS `:526`	"Adult height is typically more than 3.5 standard deviations below the mean"	Unsupported. The nearest datum in the entry (PMID:41454937) is a −3.62 height Z-score for the missense subgroup at first presentation (vs −1.99 for truncating), not adult height for all patients.
Diastrophic `:52,:82`	Finnish incidence "~1:22,000"	Unsupported; the cited snippet says only that incidence "has significantly decreased".
Pycnodysostosis `:27,:64,:874`	"37% from Europe and 31% from Asia"; "70% mature domain, 24% proregion, 6% preregion"	Correct — all figures verified present in cited PMID:21569238. Nit only: not captured in any `snippet:`, so a reader cannot trace them.

9. Mechanistic assertions in prose without evidence items

Pycnodysostosis.yaml:10-11,129-131: "osteoclasts … can acidify the resorption lacuna normally but cannot degrade the demineralized organic matrix" — the defining distinction from osteopetrosis, and the node's only evidence is "cathepsin K is a major protease in bone resorption". This is a well-established fact that deserves its own citation.
Hypophosphatasia: Craniosynostosis downstream edge (:325) has no evidence, though a supporting quote is already used elsewhere in the file.

Prevalence 3.4/million is quoted for "spondyloepiphysal dysplasia (SED)" — congenita and tarda combined — then recorded as the SEDC rate. The notes field discloses this honestly, which is good practice; flagging only so the limitation stays visible. The explanation calls PMID:31523532 "this review"; it is a case report.
Vitreous Collagen Abnormality (:290) states myopia ≈45% / retinal detachment ≈12% "of SEDC patients". Those figures are correctly evidenced elsewhere in the file (PMID:25604898) but come from a mixed 93-patient COL2A1 cohort, and the node itself carries no citation for them.

Verified-good patterns worth preserving

Several entries handle exactly these hazards well and are worth citing as models:

Achondroplasia is the strongest entry reviewed. It uses mechanistic_hypotheses to avoid asserting strict constitutive FGFR3 activation, explicitly annotates vosoritide/navepegritide as accelerated approvals on a surrogate endpoint, and flags the 5.75 cm height gain as an external-control (not randomized) comparison. It also reconciles a conflict openly: an older paper's ">90% sporadic" is quoted, with the explanation stating the structured de_novo_rate is calibrated to the current ~80%.
Thanatophoric Dysplasia Type 1 marks the TD2-equivalent mouse (PMID:9887329) as supports: PARTIAL and says why in the explanation.
Cleidocranial Dysplasia declines to assign a frequency to short stature because two cohorts disagree (71% vs 28.6%), and shows both — the correct response to conflicting frequency evidence.
Kniest Dysplasia and Pycnodysostosis transparently disclose grounding compromises (e.g. acroosteolysis mapped to the HPO hand-specific osteolysis term).
Achondroplasia correctly does not declare conforms_to the module's "Constitutive FGFR Activation" node, consistent with its own hedged receptor model, while TD1 does — the conformance edges track the biology.

Gaps noted (not errors)

Hypophosphatasia: bisphosphonates are contraindicated (they are pyrophosphate analogues and worsen the mechanism); this safety-relevant fact is not curated.
Pycnodysostosis: growth hormone therapy is not curated as a treatment even though PMID:11474477 — already cited eleven times in the entry — is titled "…linear growth after growth hormone therapy".
Cleidocranial Dysplasia: wide pubic symphysis / symphysis diastasis appears inside a quoted snippet but is not curated as a phenotype.
Achondroplasia: 30 evidence items cite a bare GeneReviews/FDA url:, which the reference validator cannot check against a cached abstract. Defensible where the quote is from chapter body text unavailable in the PubMed record, but it does mean a meaningful share of this entry's evidence is outside the automated safety net.

Suggested priority

Finding 1 (unsourced iPSC claim, 2 entries) — remove or cite.
Finding 2 (four over-claiming explanations, Hypophosphatasia) — re-quote or re-word.
Findings 3, 8 (prevalence class contradiction; unsupported figures) — mechanical fixes.
Findings 4, 5 (Campomelic mechanism transfer; MCDS superseded mechanism) — need a curator decision on framing, likely via mechanistic_hypotheses.
Findings 6, 7 (edge evidence and edge parenting) — structural cleanup.