some bacdive chemical mappings still from custom_curies.yaml

[realmarcin]

Two Sources of Production Edges in BacDive

1. 8,616 edges with METPO:2000202 (from BacDive API metabolite production assays)

Source: BacDive API data with specific metabolite production results (yes/no)

kgmicrobe.strain:bacdive_10422 → METPO:2000202 → CHEBI:16136 (L-lactate)

Flow:

• BacDive API provides: {"metabolite": "L-lactate", "chebi_id": "16136", "production": "yes"}
• Code maps production="yes" → METPO:2000202 via load_metpo_metabolite_production_mappings()
• Creates edge: strain → METPO:2000202 → specific ChEBI compound

2. 850 edges with biolink:produces (from custom_curies.yaml category mappings)

Source: BacDive API keyword data like "amino acid production", "pigmented", "toxin production"

kgmicrobe.strain:bacdive_X → biolink:produces → CHEBI:36047 (antibiotic class)

Flow:

• BacDive API keywords contain traits like "amino acid production"
• metabolite_mapping.json maps these to ChEBI chemical classes:
  • "CHEBI:36047": "antibiotic_compound" (508 edges)
  • "CHEBI:26130": "pigmented" (166 edges)
  • "CHEBI:27026": "toxin" (53 edges)
  • etc.
• custom_curies.yaml defined the predicate as biolink:produces
• Creates edge: strain → biolink:produces → ChEBI class