Dismech Ontology Terms Skill

Overview

Add and validate ontology term references in the dismech disorder knowledge base. This ensures phenotypes, cell types, biological processes, and other entities are properly linked to authoritative ontology terms with correct IDs and labels.

When to Use

• Adding phenotype_term to phenotype entries (uses HP - Human Phenotype Ontology)
• Adding term to cell_types entries (uses CL - Cell Ontology)
• Adding term to biological_processes entries (uses GO - Gene Ontology)
• Adding disease_term to disease entries (uses MONDO)
• Validating existing ontology term references
• Fixing label mismatches between preferred_term and ontology labels

Term Object Structure

All term references follow this YAML structure:

yaml
1# For phenotypes:
2phenotype_term:
3  preferred_term: <Human readable name>
4  term:
5    id: HP:XXXXXXX
6    label: <Exact HP label from ontology>
7
8# For cell types:
9cell_types:
10- preferred_term: <Human readable name>
11  term:
12    id: CL:XXXXXXX
13    label: <Exact CL label from ontology>
14
15# For biological processes:
16biological_processes:
17- preferred_term: <Human readable name>
18  term:
19    id: GO:XXXXXXX
20    label: <Exact GO label from ontology>

Ontology Lookup with OAK

Use the Ontology Access Kit (OAK) to look up terms:

Exact Match

bash
1uv run runoak -i sqlite:obo:hp info "seizure"
2# Returns: HP:0001250 ! Seizure

Fuzzy Search

bash
1uv run runoak -i sqlite:obo:hp info "l~cognitive impairment"
2# Returns multiple matches - select the most appropriate

Get Full Term Details

bash
1uv run runoak -i sqlite:obo:cl info CL:0000540 -O obo
2# Returns complete term information including definition

Common Ontology Prefixes

Specificity Guidelines

Critical: Always use the most specific term that accurately describes the entity:

When a fuzzy search returns multiple results:

1. Review all candidates
2. Check term definitions with -O obo flag
3. Select the term that most precisely matches the biological context
4. If no specific term exists, use the closest parent but note the limitation

Validation

After adding terms, validate with:

bash
1just validate-terms

This checks:

• Term IDs exist in the ontology
• Labels match the canonical ontology labels exactly
• Required fields are present

Fixing Label Mismatches

If validation reports a label mismatch:

LABEL MISMATCH: Cholera.yaml
  Term: HP:0003394
  Expected: Muscle cramps
  Actual: Muscle spasm

Update the label field to match the ontology's canonical label exactly.

Batch Processing

To find entries missing term annotations:

python
1import yaml
2import glob
3
4for f in glob.glob("kb/disorders/*.yaml"):
5    with open(f) as file:
6        data = yaml.safe_load(file)
7    for pheno in data.get('phenotypes', []):
8        if 'phenotype_term' not in pheno:
9            print(f"{f}: {pheno.get('name')} - missing phenotype_term")

Common Patterns

Adding HPO to a Phenotype

1. Look up term: uv run runoak -i sqlite:obo:hp info "l~<phenotype name>"
2. Verify specificity: uv run runoak -i sqlite:obo:hp info <HP:ID> -O obo
3. Add to YAML:

   yaml
   1phenotype_term:
   2  preferred_term: <Original Name>
   3  term:
   4    id: <HP:ID>
   5    label: <Exact label from OAK>

4. Validate: just validate-terms

Adding CL to Cell Types

1. Look up term: uv run runoak -i sqlite:obo:cl info "l~<cell type>"
2. Verify specificity
3. Add term: block under the cell_type entry
4. Validate