sdrf-terms

name	sdrf:terms
description	Use when the user needs to find, verify, or compare ontology terms for SDRF columns. Triggers on questions about ontology terms, accessions, or which term to use.
user-invocable	true
argument-hint	[column name] [search term]

SDRF Ontology Term Lookup

You are helping the user find the correct ontology term for an SDRF column.

Step 1: Identify the Column and Ontology

Read spec/sdrf-proteomics/TERMS.tsv and find the row for the column the user is asking about. The values field tells you which ontology(ies) to search.

Examples from TERMS.tsv:

• organism → values: NCBITaxon → search OLS with ontologyId ncbitaxon
• disease → values: MONDO, EFO, DOID, PATO → search these ontologies
• organism part → values: UBERON, BTO → search UBERON first, BTO as fallback
• cell type → values: CL, BTO → search CL first, BTO as fallback
• instrument → values: MS → search MS ontology
• modification parameters → values: UNIMOD → use UNIMOD accessions
• cleavage agent details → values: MS → search MS ontology

Always read TERMS.tsv rather than relying on memorized ontology mappings — the spec may add new columns or change ontology sources.

Step 2: Search OLS

Use the OLS MCP tools to find the term:

Primary search:
  mcp OLS → searchClasses(query="<user term>", ontologyId="<ontology>")

If no results or too many:
  mcp OLS → search(query="<user term>")
  Filter results to the correct ontology manually

For broader semantic search:
  mcp OLS → searchClassesWithEmbeddingModel(query="<description>", model="<model>")
  (Call listEmbeddingModels first to get available models with can_embed=true)

Step 3: Evaluate Specificity

When presenting results, assess specificity:

Too Generic (suggest more specific)

• "cancer" → suggest "breast carcinoma", "lung adenocarcinoma", etc.
• "tissue" → suggest the actual tissue name
• "cell" → suggest the actual cell type
• "brain" might be OK, but "temporal cortex" is better if known

Appropriately Specific

• "breast carcinoma" (EFO:0000305) — good for a breast cancer study
• "liver" (UBERON:0002107) — good for tissue-level studies
• "T cell" (CL:0000084) — good if subtype unknown

Too Specific (might be too narrow)

• "left breast upper inner quadrant" — probably too specific for most studies

To check specificity, use hierarchy navigation:

mcp OLS → getAncestors(ontologyId="<ont>", classIri="<iri>")
mcp OLS → getChildren(ontologyId="<ont>", classIri="<iri>")

Step 4: Cross-Ontology Mapping

When the user has a term from one ontology but needs another:

Example: User has DOID term, needs EFO equivalent
  1. Get the DOID term details: mcp OLS → fetch(id="doid+<iri>")
  2. Search EFO for the same concept: mcp OLS → searchClasses(query="<label>", ontologyId="efo")
  3. Present both options with accessions

For disease terms, SDRF accepts EFO, MONDO, or DOID. Recommend:

• EFO as first choice (most commonly used in SDRF)
• MONDO as second choice (good cross-references)
• DOID as third choice

Step 5: Present Results

For each term found, present:

Term: breast carcinoma
Accession: EFO:0000305
Ontology: Experimental Factor Ontology (EFO)
Definition: A carcinoma that arises in the breast region.
Synonyms: breast cancer, mammary carcinoma
Parent: carcinoma (EFO:0000228)
SDRF format: breast carcinoma
Column: characteristics[disease]

Alternative terms:
  - invasive breast carcinoma (EFO:0010132) — more specific, if applicable
  - breast ductal carcinoma (EFO:0000298) — subtype-specific

Special Cases

"Normal" / "Healthy" / "Control"

• For disease: use normal with accession PATO:0000461
• Do NOT use: "healthy", "control", "none", "N/A"

"Not Available" vs "Not Applicable"

• not available — the information exists but wasn't captured
• not applicable — the property doesn't apply (e.g., cell line for a tissue sample)
• Check TERMS.tsv allow_not_available and allow_not_applicable for the specific column

Cell Lines

• Use the Cellosaurus database (https://www.cellosaurus.org/) for cell line identification
• SDRF uses three columns for cell lines:
  • characteristics[cell line] — name from CLO, BTO, or EFO ontology
  • characteristics[cellosaurus accession] — format: CVCL_XXXX (e.g., CVCL_0030 for HeLa)
  • characteristics[cellosaurus name] — official Cellosaurus name
• Common examples: HeLa (CVCL_0030), HEK293 (CVCL_0045), MCF7 (CVCL_0031), A549 (CVCL_0023)
• To find a Cellosaurus accession: search https://www.cellosaurus.org/search (not OLS)
• Cross-reference Cellosaurus for: species of origin, disease, tissue of origin, STR profile

Instruments

• Format in SDRF: AC=MS:1001911;NT=Q Exactive HF
• Search MS ontology for the instrument model
• Include manufacturer in search if needed

Modifications

• ALWAYS use UNIMOD accessions, not PSI-MOD
• The format is: NT=<name>;AC=UNIMOD:<id>;TA=<amino acid>;MT=<Fixed|Variable>
• Double-check the UNIMOD:1/UNIMOD:21 swap (Acetyl vs Phospho)