Skip to content

Nemaline Myopathy Curation Interaction Summary

Date: 2026-02-19 Branch: nemaline_rod_myopathy Deep Research Provider: Falcon


1. Primary Request and Intent

The user first asked for a tour of the dismech project (a LinkML-based disease pathophysiology knowledge base). After the walkthrough, the user invoked /curate nemaline rod myopathy to create a new disorder entry following the initiate-new-disorder-creation skill workflow.

After the initial curation was complete, the user challenged whether 7 specific foundational papers were considered during curation — they were not, as the falcon deep research provider focused on 2023–2024 literature. The user then requested incorporation of those 7 papers into the entry, covering the full historical breadth of NM literature from 1984 to 2024.

2. Key Technical Concepts

  • LinkML schema validation: Disorder YAML files must conform to the Disease class in src/dismech/schema/dismech.yaml
  • Three-layer validation stack: schema validation (just validate), ontology term validation (just validate-terms), and reference/snippet validation (just validate-references)
  • Evidence-backed curation: Every factual claim requires a PMID reference with an exact quoted snippet from the abstract — no paraphrasing allowed
  • Ontology term bindings: Phenotypes → HPO, Cell types → CL, Biological processes → GO, Treatments → MAXO, Diseases → MONDO
  • OAK (Ontology Access Kit): Used for looking up ontology terms via uv run runoak -i sqlite:obo:<ontology> info
  • Deep research via falcon: just research-disorder falcon Nemaline_Myopathy runs external deep research and outputs to research/Nemaline_Myopathy-deep-research-falcon.md
  • Reference fetching: just fetch-reference PMID:nnnnn or just fetch-reference DOI:... caches abstracts to references_cache/
  • Pre-write validation hook: A git hook at .claude/hooks/validate_disorder_hook.py blocks writes that would create invalid YAML files
  • Compliance scoring: just compliance reports coverage percentages for evidence, terms, and other recommended fields

3. Workflow Steps

Step 1: Verify Disorder Doesn't Exist

  • Checked kb/disorders/ — no existing Nemaline Myopathy file found
  • Looked up MONDO term: MONDO:0018958 (nemaline myopathy)

Step 2: Create Initial YAML Scaffold

  • Created kb/disorders/Nemaline_Myopathy.yaml with textbook knowledge
  • Included basic structure: name, category (Mendelian), description, disease_term, parents, subtypes, pathophysiology, phenotypes, genetics, treatments

Step 3: Deep Research

  • Ran just research-disorder falcon Nemaline_Myopathy
  • Research output written to research/Nemaline_Myopathy-deep-research-falcon.md
  • Falcon returned 42 citations focused on 2023–2024 literature
  • Key limitation: falcon's recency bias missed foundational/classic NM papers

Step 4: DOI-to-PMID Resolution

  • Falcon output provided DOIs rather than PMIDs
  • NCBI PMC ID converter API returned 404 errors for all DOIs
  • Solved by using Biopython Entrez.esearch to search PubMed by title keywords
  • Successfully resolved 10 DOIs to correct PMIDs

Step 5: Evidence Enhancement

  • Fetched and read 10 abstracts from falcon research
  • Added exact snippet quotes from abstracts to evidence items
  • Added ontology terms (HP, CL, GO, MAXO) via OAK lookups

Step 6: User-Identified Paper Incorporation

  • User identified 7 foundational papers not covered by falcon research:
  • PMID:17272906 — Indian cohort study (2007)
  • PMID:31228046 — Comprehensive NM review (2019)
  • PMID:9372751 — Adult-onset NM case report (1998)
  • PMID:35810298 — Severe ACTA1-NM study (2022)
  • PMID:6481414 — Early adult NM report with mixed pathology (1984)
  • PMID:19208402 — NM with isolated respiratory failure (2009)
  • PMID:39691005 — SLONM case with monoclonal protein (2024)
  • All 7 papers were fetched, read, and incorporated with exact snippets across multiple sections

Step 7: Validation

  • just validate — schema validation passed
  • just validate-terms-file — ontology term validation passed
  • just validate-references — reference/snippet validation passed
  • just compliance — 82.4% compliance (82.8% weighted)

4. Files Created/Modified

File Action Description
kb/disorders/Nemaline_Myopathy.yaml Created Main disorder entry (711 lines)
research/Nemaline_Myopathy-deep-research-falcon.md Generated Falcon deep research output
pages/disorders/Nemaline_Myopathy.html Generated Rendered HTML page
references_cache/PMID_38634969.md Fetched Karimi 2024 — NEB thin filament mechanisms
references_cache/PMID_36703211.md Fetched Nicolau 2023 — Nemaline rod signatures (iNM vs SLONM)
references_cache/PMID_38376469.md Fetched Galli 2024 — Tirasemtiv in NEM3 mouse model
references_cache/PMID_36661122.md Fetched Casey 2023 — NRAP in nebulin-deficient NM
references_cache/PMID_36960434.md Fetched 2023 — Pediatric NM systematic review
references_cache/PMID_36714460.md Fetched Moreno 2023 — NEB-NM cohort (33 patients)
references_cache/PMID_39651462.md Fetched van Kleef 2024 — NEM6/KBTBD13 (24 patients)
references_cache/PMID_37432316.md Fetched Mansur 2023 — KLHL40/UPS mechanism
references_cache/PMID_38500810.md Fetched 2024 — ACTA1 case report
references_cache/PMID_39501809.md Fetched 2024 — Additional NM reference
references_cache/PMID_17272906.md Fetched Jain 2007 — Indian NM cohort
references_cache/PMID_31228046.md Fetched Ravenscroft 2019 — Authoritative NM review
references_cache/PMID_9372751.md Fetched Gyure 1997 — Adult-onset NM with HIV
references_cache/PMID_35810298.md Fetched Néstor-Bergmann 2022 — Severe ACTA1-NM
references_cache/PMID_6481414.md Fetched Engel 1984 — Early adult NM report
references_cache/PMID_19208402.md Fetched Kelly 2009 — NM respiratory failure
references_cache/PMID_39691005.md Fetched Vashisht 2024 — SLONM case

5. Errors and Fixes

Schema Validation Errors (caught by pre-write hook)

Error Root Cause Fix
description not allowed in Genetic class Schema uses notes for Genetic, not description Changed all description: to notes: in genetic entries
notes not allowed in Inheritance class Overcorrection from previous error; schema allows description here Changed back to description: in inheritance entries
percentage and population not valid in EpidemiologyInfo Used Prevalence-style fields not in schema Restructured to name: Incidence with description and notes
Snippet validation failure for PMID:39651462 Brackets [4.9-8.1] in snippet may have caused matching issues Replaced with different snippet from same abstract

DOI Resolution Errors

Error Root Cause Fix
NCBI PMC ID converter API 404 API endpoint returned 404 for all DOIs Used Biopython Entrez.esearch with title keywords
Guessed PMIDs were wrong Manual PMID guessing returned unrelated papers (neurofilaments, breast cancer, scorpion venom) Abandoned guessing; used systematic title-based search

Deep Research Limitation

Issue Impact Mitigation
Falcon provider focused on 2023–2024 Missed 7 foundational papers (1984–2022) User identified gaps; papers manually incorporated

6. Final Entry Statistics

Metric Value
Subtypes 5 (Severe Congenital, Typical Congenital, Childhood-Onset, Adult-Onset, NEM6)
Pathophysiology mechanisms 6 (Thin Filament Dysfunction, Nemaline Rod Formation, Thin Filament Length Dysregulation, UPS/KLHL40, Nuclear Envelope Disruption, NRAP-Mediated Sarcomere Disorganization)
Phenotypes 9
Genetic entries 10 (NEB, ACTA1, KBTBD13, KLHL40, TPM3, TPM2, TNNT1, CFL2, LMOD3, KLHL41)
Treatments 7 (including 2 investigational)
Epidemiology entries 1
Unique PMIDs 17 (spanning 1984–2024)
Compliance 82.4% (82.8% weighted)
Validation All 3 layers pass (schema, terms, references)

7. PMIDs Used

PMID First Author Year Key Contribution
38634969 Karimi 2024 NEB thin filament length mechanisms, omecamtiv mecarbil
36703211 Nicolau 2023 Rod signatures: iNM vs SLONM
38376469 Galli 2024 Tirasemtiv in NEM3 mouse model
36661122 Casey 2023 NRAP as disease modifier in NEB-NM
36960434 2023 Pediatric NM systematic review
36714460 Moreno 2023 NEB-NM cohort (33 patients)
39651462 van Kleef 2024 NEM6/KBTBD13 cross-sectional study (24 patients)
37432316 Mansur 2023 KLHL40/UPS/ER-Golgi trafficking mechanism
38500810 2024 Novel ACTA1 variant case report
39501809 2024 Additional NM reference
17272906 Jain 2007 Indian NM cohort, epidemiology (0.53% of muscle diseases)
31228046 Ravenscroft 2019 Comprehensive NM review (authoritative)
9372751 Gyure 1997 Adult-onset NM with HIV, rod morphology description
35810298 Néstor-Bergmann 2022 Severe ACTA1-NM, nuclear envelope disruption
6481414 Engel 1984 Early adult NM case, mixed pathology
19208402 Kelly 2009 NM with isolated respiratory failure
39691005 Vashisht 2024 SLONM case with monoclonal protein

8. Lessons Learned

  1. Deep research providers have recency bias: Falcon focused on 2023–2024 papers, missing foundational literature. Supplementary manual reference searching is needed, especially for well-established diseases with decades of published literature.

  2. Schema field constraints vary by class: Different LinkML classes (Genetic, Inheritance, EpidemiologyInfo) have different allowed slots. Always check the schema before writing YAML.

  3. DOI-to-PMID resolution is fragile: The NCBI PMC ID converter API can fail. Title-based PubMed search via Biopython Entrez.esearch is a reliable fallback.

  4. Never guess PMIDs: Attempting to guess PMIDs based on author/year produces completely unrelated papers. Always use systematic search.

  5. Pre-write validation hooks catch errors early: The .claude/hooks/validate_disorder_hook.py hook prevented multiple schema-invalid writes from being saved, forcing immediate correction.

  6. Snippet bracket characters can cause validation issues: Brackets in quoted snippets (e.g., [4.9-8.1]) may not survive the reference validation pipeline. Use alternative snippets when possible.

9. User Messages (Chronological)

  1. "Give me a tour of the dismech project"
  2. /curate nemaline rod myopathy
  3. "is there evidence in the output, that you considered and read these papers: https://pubmed.ncbi.nlm.nih.gov/17272906/ https://pubmed.ncbi.nlm.nih.gov/31228046/ https://pubmed.ncbi.nlm.nih.gov/9372751/ https://pubmed.ncbi.nlm.nih.gov/35810298/ https://pubmed.ncbi.nlm.nih.gov/6481414/ https://pubmed.ncbi.nlm.nih.gov/19208402/ https://pubmed.ncbi.nlm.nih.gov/39691005/"
  4. "yes" (confirming to incorporate the 7 papers)
  5. "write out this interaction to a md file"