embl-ebi-ols - SKILL.md Agent Skill

name: embl-ebi-ols description: > Query and search the EMBL-EBI Ontology Lookup Service (OLS) for biomedical ontology terms, definitions, and hierarchies across 250+ ontologies (e.g., GO, DOID, HP). Use when the user asks to search for terms, retrieve details, navigate hierarchies (parents, children, ancestors), look up properties and individuals, get autocomplete suggestions, or access ontology metadata and statistics.

EMBL-EBI Ontology Lookup Service (OLS)

Prerequisites

uv: Read the uv skill and follow its Setup instructions to ensure uv is installed and on PATH.
User Notification: If LICENSE_NOTIFICATION.txt does not already exist in this skill directory then (1) prominently notify the user to check the terms at https://www.ebi.ac.uk/ols4/api-docs, then (2) create the file recording the notification text and timestamp.

Core Rules

[!IMPORTANT] Use the Utility Scripts: You MUST ALWAYS use the provided utility script under scripts/ for all API interactions, including checking status. NEVER use curl or custom Python requests to query API directly.
Rate Limiting & Resilience: You MUST respect EBI's Terms of Use with a maximum 5 requests per second. The provided utility scripts automatically enforce this.
Notification: If this skill is used, ensure this is mentioned in the output.

When to Use — Quick Recipes

Use this skill whenever a user query matches one of these patterns:

Definition of a disease, phenotype, or term → get_term.py --obo_id <ID> --summary
Subtypes or children of a term → get_term.py --obo_id <ID> --relations children
Parent of a term → get_term.py --obo_id <ID> --relations parents
Ancestors / disease categories / classified under → get_term.py --obo_id <ID> --relations ancestors
Root terms of an ontology → get_term.py --ontology <id> --roots
Hierarchical parents (is-a + part-of) → get_term.py --obo_id <ID> --relations hierarchicalParents
Structures part of / hierarchical children → get_term.py --obo_id <ID> --relations hierarchicalChildren
Compare direct vs hierarchical parents → get_term.py --obo_id <ID> --relations parents,hierarchicalParents
Search for a term (e.g., "apoptosis" in GO) → search_ols.py --query "..." --ontology <id>
Find a GO term matching a function → search_ols.py --query "..." --ontology go --exact
Search in MONDO, CHEBI, CL, UBERON → search_ols.py --query "..." --ontology <id> --defining
Paginate search results / next page → search_ols.py --query "..." --rows N --start <offset>
Autocomplete a partial name → suggest_ols.py --query "..."
Ontology metadata (e.g., EFO info) → get_ontology.py --id <id>
OLS index statistics → get_stats.py

Multi-step queries (e.g., "What is the parent of myocardial infarction?"): When the user names a term but you don't know its OBO ID, complete in exactly 2 steps — do NOT search across multiple ontologies:

Search in the single most appropriate ontology: search_ols.py --query "myocardial infarction" --ontology doid --exact --rows 1 --output /tmp/step1.json

Get relations using the OBO ID from step 1: get_term.py --obo_id DOID:5844 --relations parents --output /tmp/step2.json

Ontology selection rule: ALWAYS use doid for common human diseases (e.g., diabetes, cancer), hp for phenotypes, go for gene functions, chebi for chemicals, uberon for anatomy, cl for cell types. Use mondo ONLY when cross-species context is explicitly mentioned or needed.

Utility Scripts

1. Search Terms Across Ontologies

Search for ontology terms by keyword and return clean JSON.

uv run scripts/search_ols.py --query "diabetes" \
  --rows 5 --output /tmp/ols_search_results.json 2>/dev/null

Important: --output is required for all scripts. Results are always written to the specified file. For larger output, you can limit --rows (e.g., 5-10) or paginate using --start.

Returned Fields: JSON results include iri, label, description, ontology_name, ontology_prefix, obo_id, short_form, type, is_defining_ontology, and exact_synonyms.

Pagination: Output includes a pagination block with start, rows, and has_more so you can decide whether to fetch more results.

Options:

--query: Search string (required). Searches labels, synonyms, descriptions, and identifiers.
--ontology: Filter by ontology ID (e.g., go, doid, efo, hp). Recommended when you know which ontology to search — avoids noise from 250+ ontologies.
--type: Filter by entity type: class, property, individual, or ontology.
--exact: Flag for exact label match only. Use this for entity resolution when mapping a user's string to a specific ontology term ID.
--defining: Only return terms from their defining (authoritative) ontology. E.g., GO:0005634 only from GO, not cross-referenced copies.
--obsolete: Flag to include obsolete terms in results.
--local: Only return terms in their defining ontology.
--childrenOf: Restrict to children of given term IRI(s), comma-separated.
--allChildrenOf: Restrict to all children including transitive relations (part of, develops from), comma-separated IRIs.
--queryFields: Comma-separated fields to search in (e.g., label,synonym,description).
--fieldList: Comma-separated fields to return.
--groupField: Group results by unique IRI.
--isLeaf: Only return leaf terms (no children).
--rows: Number of results to return (default 10).
--start: Pagination offset (default 0).
--output: File path to save results (required).

2. Autocomplete / Suggest

Get autocomplete suggestions for partial term names.

uv run scripts/suggest_ols.py --query "diabet" --rows 5 \
  --output /tmp/ols_suggest.json 2>/dev/null

Options:

--query: Partial term to autocomplete (required).
--ontology: Filter by ontology ID(s), comma-separated.
--rows: Number of suggestions (default 10).
--start: Pagination offset (default 0).
--output: File path to save results (default: stdout).

3. Get Term Details

Retrieve full details for a specific ontology term by its OBO ID or IRI.

uv run scripts/get_term.py --obo_id "GO:0005634" \
  --output /tmp/ols_term.json 2>/dev/null

Returned Fields: JSON includes iri, label, description, obo_id, synonyms, ontology_name, is_obsolete, is_defining_ontology, has_children, is_root, annotation, in_subset, and any requested relations.

Summary Mode: Use --summary to get a clean, human-readable block on stdout (Label, OBO ID, Ontology, Definition, Synonyms). The full JSON is always saved to the --output file.

uv run scripts/get_term.py --obo_id "GO:0005634" --summary \
  --output /tmp/nucleus_full.json

Options:

--obo_id: OBO-style identifier (e.g., GO:0005634, DOID:9351). Mutually exclusive with --iri. Auto-converts to IRI with double encoding.
--iri: Full IRI of the term. Mutually exclusive with --obo_id.
--ontology: Ontology ID (auto-derived from --obo_id if not provided).
--relations: Comma-separated list of relations to fetch.
- Direct (is-a only): parents, children, ancestors, descendants
- Hierarchical (is-a + transitive like "part of", "develops from"): hierarchicalParents, hierarchicalChildren, hierarchicalAncestors, hierarchicalDescendants
- Graph: graph — full graph JSON for a term
Note: Use hierarchical variants for anatomical/developmental ontologies (UBERON, CL) where transitive relations like "part of" and "develops from" are critical for navigating the hierarchy.
--roots: List root terms of the ontology (requires --ontology).
--preferred_roots: List preferred root terms (requires --ontology).
--summary: Human-readable summary on stdout, full JSON to --output.
--output: File path to save results (default: stdout).

4. Get Property Details

Retrieve details for an ontology property (relation type) with hierarchy.

uv run scripts/get_property.py --obo_id "BFO:0000051" --ontology go \
  --output /tmp/ols_property.json 2>/dev/null

Options:

--obo_id: OBO-style ID of the property. Mutually exclusive with --iri.
--iri: Full IRI of the property. Mutually exclusive with --obo_id.
--ontology: Ontology ID (required with --iri).
--relations: Comma-separated: parents, children, ancestors, descendants.
--roots: List root properties of the ontology (requires --ontology).
--output: File path to save results (default: stdout).

5. Get Individual Details

Retrieve details for an ontology individual (instance).

uv run scripts/get_individual.py --obo_id "IAO:0000103" --ontology iao --types \
  --output /tmp/ols_individual.json 2>/dev/null

Options:

--obo_id: OBO-style ID. Mutually exclusive with --iri.
--iri: Full IRI. Mutually exclusive with --obo_id.
--ontology: Ontology ID (required with --iri).
--types: Fetch the direct types (classes) of this individual.
--alltypes: Fetch all types including ancestor classes.
--output: File path to save results (default: stdout).

6. Get Ontology Information

List available ontologies or retrieve details for a specific one.

uv run scripts/get_ontology.py --id go \
  --output /tmp/ols_ontology.json 2>/dev/null

Options:

--id: Specific ontology ID (e.g., go, efo, doid). If omitted, lists all ontologies.
--page: Page number for pagination (default 0).
--size: Number of ontologies per page (default 20).
--output: File path to save results (default: stdout).

7. Get OLS Statistics

Retrieve index statistics (total ontologies, classes, properties, individuals).

uv run scripts/get_stats.py --output /tmp/ols_stats.json 2>/dev/null

Options:

--output: File path to save results (default: stdout).

Reference

API Reference: See references/api_reference.md for common ontology IDs, OBO ID format, and key API endpoints.

Workflow

Use suggest_ols.py for autocomplete when you have a partial term name.
Search for terms using search_ols.py. Use --defining to prioritize authoritative definitions. Use --exact for entity resolution.
If full details are needed, use get_term.py with the OBO ID or IRI. Use --summary for a concise view.
To explore a term's hierarchy, use get_term.py --relations parents,children for is-a only, or --relations hierarchicalParents,hierarchicalChildren for "part of" etc.
To explore from the top down, use get_term.py --ontology go --roots.
For properties or individuals, use get_property.py or get_individual.py.
To discover available ontologies, use get_ontology.py.
To check OLS index status, use get_stats.py.