# OAK Configuration for ai-gene-review term validation
# Maps ontology prefixes to OAK adapter strings
#
# Adapter types:
#   ols:<name>         - EBI Ontology Lookup Service (online, no local download)
#   sqlite:obo:<name>  - Pre-built SQLite databases (fast, offline)
#   null               - Skip validation for this prefix
#
# Most OBO ontologies use the OLS adapter (ols:<slug>) so validation queries
# the EBI Ontology Lookup Service directly instead of downloading large SQLite
# builds. GO stays on sqlite:obo:go: it drives the dynamic-enum / branch
# reachability checks (molecular_function, in_complex, locations, ...), and the
# OLS adapter's ancestor traversal is unreliable for that (it raises on obsolete
# / detached terms and can return incomplete closures). RHEA is not hosted in
# OLS, so it also stays on sqlite:obo:rhea.
#
# Note: ONLY prefixes listed here are validated. Unlisted prefixes are skipped
# with a warning. Use null to explicitly skip without warning.

ontology_adapters:
  # Non-ontology prefixes that should be skipped
  TEMP: null        # Temporary/placeholder identifiers
  PMID: null        # PubMed IDs
  PMC: null         # PubMed Central IDs
  DOI: null         # Digital Object Identifiers
  ISBN: null        # Book identifiers
  ISSN: null        # Journal identifiers
  UniProt: null     # UniProt accessions
  UniProtKB: null   # UniProt Knowledgebase accessions
  FB: null          # FlyBase gene (FBgn) and gene group (FBgg) ids
  RefSeq: null      # RefSeq accessions
  ENSEMBL: null     # Ensembl IDs
  EC: null          # Enzyme Commission numbers
  PDB: null         # Protein Data Bank IDs
  OMIM: null        # Online Mendelian Inheritance in Man
  file: null        # File path references
  GO_REF: null      # Gene Ontology reference collection
  VZ: null          # ViralZone (ExPASy) page ids, e.g. VZ:1951 - this is GO's own
                    # dbxref spelling for ViralZone (152 GO terms carry one), not the
                    # bioregistry 'viralzone' prefix. Deliberately unvalidated: pages
                    # are prose with no local cache and no ontology to resolve against.
  Reactome: null    # Reactome pathway database
  skos: null        # SKOS mapping predicates (e.g. skos:relatedMatch) - not ontology-validated
  semapv: null      # SSSOM mapping-justification vocabulary - not ontology-validated
  RO: null          # Relation Ontology predicates (e.g. RO:0002327 enables) - not label-validated here

  # Standard OBO ontology prefixes - use the OLS adapter (ols:<slug>)
  GO: sqlite:obo:go  # stays on SQLite: needed for dynamic-enum branch reachability
  ARO: ols:aro  # Antibiotic Resistance Ontology (CARD) - for ARO->GO mapping validation
  CHEBI: ols:chebi
  UBERON: ols:uberon
  CL: ols:cl
  HP: ols:hp
  MONDO: ols:mondo
  PR: ols:pr
  SO: ols:so
  PATO: ols:pato
  NCBITaxon: ols:ncbitaxon
  RHEA: sqlite:obo:rhea  # RHEA is not hosted in OLS; keep the SQLite build
  PO: ols:po  # Plant Ontology - anatomical locations in plant modules (e.g. root)
  SBO: null  # Systems Biology Ontology (module regulation predicates) - no working
             # offline sqlite:obo build (labels resolve to None), so not label-validated here
  MetaCyc: null  # MetaCyc pathway/reaction ids - not OBO label-validated here
  MIBiG: null  # Minimum Info about a Biosynthetic Gene cluster - not label-validated here
  RFAM: null  # Rfam RNA family ids (e.g. RFAM:RF00169) used as module RNA family selectors
  NCBIGene: null  # NCBI Gene ids, used for noncoding RNA genes that have no UniProt entry
  NCBIfam: null  # NCBIfam/TIGRFAM family ids (e.g. NCBIfam:TIGR01356) used as module
                 # family selectors - not OBO label-validated here
  InterPro: null  # InterPro entry ids (e.g. InterPro:IPR053409) - not OBO label-validated here
  # PANTHER family/subfamily ids (PTHR12345, PTHR12345:SF7) are validated against
  # a local OBO built from PANTHER's own HMM classifications (see
  # ai_gene_review.etl.panther_families; rebuild with `just build-panther-obo`).
  # PANTHER publishes no term-lookup service and InterPro indexes only families
  # -- it returns 204 for every :SF accession -- so this local artifact is the
  # only way to check subfamily ids at all. The path is resolved relative to the
  # repository root. PTN ancestral-node ids share this prefix but are NOT routed
  # here: they carry stronger PAINT/IBD evidence checks (see validate_paint_ptns).
  PANTHER: simpleobo:interpro/panther/panther.obo

# Reviewed labels from the authoritative live service that are newer than the
# sqlite:obo:go snapshot. Remove entries once the snapshot exposes each label
# as its primary label or a synonym.
term_label_aliases:
  GO:0008883:
    - glutamyl-tRNA reductase (NADP+) activity
  GO:0070819:
    - protoporphyrinogen oxidase activity, quinone as acceptor
  GO:0004729:
    - protoporphyrinogen oxidase activity, oxygen as acceptor