| name | kg-query |
| description | Query KG-Microbe for organism information, growth media preferences, and metabolic capabilities |
KG-Microbe Query Skill
Query the KG-Microbe knowledge graph for organism information including taxonomy, phenotypic traits, growth media, and media composition.
Usage
Basic Query
poetry run kg query-organism "Eggerthella lenta"
Save to File
poetry run kg query-organism "Corynebacterium glutamicum" -o report.md
Options
--db-path PATH: Custom database location (default: data/merged/kg-microbe.duckdb)--force-reload: Rebuild database from TSV files--nodes-path PATH: Custom nodes.tsv location--edges-path PATH: Custom edges.tsv location--output FILE: Save report to file instead of printing
Query Capabilities
- Name Resolution: Fuzzy matching on organism names and synonyms
- Taxonomy: Species classification hierarchy
- Phenotypic Traits: Oxygen preference, metabolism, morphology, Gram stain
- Media Preferences: Growth media edges use a Biolink predicate (e.g.
biolink:located_in) and carry the METPO semantics in the relation column — METPO:2000517 (grows in) and METPO:2000518 (doesn't grow in). Filter on relation when querying.
- Media Composition: 2-hop traversal (organism → media → solutions → chemicals)
- Strain Information: All strain records linked to species
Data Sources
- Nodes: 1.5M entries (883K organism taxa)
- Edges: 6.1M relationships
- Primary sources: BacDive, MediaDive, MadinEtal, BactoTraits, GTDB, UniProt
Example Organisms
- Eggerthella lenta (NCBITaxon:84112) - 65+ strains, 4 growth media
- Corynebacterium glutamicum (NCBITaxon:1718)
- Escherichia coli (NCBITaxon:562)
- Bacillus subtilis (NCBITaxon:1423)
Performance
- Initial database load: ~60 seconds (one-time, cached)
- Subsequent queries: <1 second
- Database file size: ~800MB
Technical Details
- Database: DuckDB persistent storage
- Query engine: SQL with 1-hop and 2-hop graph traversal
- Output format: Markdown report with structured sections
- Caching: Database persisted to disk for fast repeat queries
