# How each ProtNLM2 benchmark cohort was drawn from the released predictions.
# Curated wording only; every count is derived by build_benchmark_funnel.py.
# `pool` names the released set the cohort was drawn from:
#   pilot          the whole published pilot list
#   taxon:<id>     pilot records of one NCBI taxon
#   mod_species    pilot records of the MOD_EVOLUTION20 species
#   pombe_export   S. pombe entries from the pre-release export (not on the list)
# `eligible` (default go_or_function) says which pool records the cohort could
# be drawn from: go_or_function, all (the whole pool), complement (records
# without GO/function text), or none (the tier mixes listed and off-list records).
# Paired reference cohorts (e.g. HORSE40_HUMAN_PAIR) are not drawn from the
# released predictions and are listed under `not_drawn_from_release`.
cohorts:
  - id: ARGO50
    label: ARGO-50
    drawn_from: whole pilot, 14 taxon groups
    pool: pilot
    method: hand-picked
    detail: stratified by taxon group and output type, incl. 9 names-only
    eligible: all
  - id: HORSE40
    label: HORSE40
    drawn_from: Equus caballus
    pool: taxon:9796
    method: hand-picked
    detail: informative claims and biological diversity
  - id: FLY41
    label: FLY41
    drawn_from: Drosophila melanogaster
    pool: taxon:7227
    method: census
    detail: every record with GO or function text
  - id: FLY_LOCATION_KEYWORD
    label: FLY location/keyword tier
    drawn_from: Drosophila melanogaster
    pool: taxon:7227
    method: census
    detail: every record with only location or keyword output; queued
    eligible: complement
  - id: FLY_NEXT20
    label: FLY_NEXT20
    drawn_from: Drosophila melanogaster
    pool: taxon:7227
    method: hand-picked
    detail: metabolic enzymes and transport; 9 off the list; queued
    eligible: none
  - id: NEUROSPORA20
    label: NEUROSPORA20
    drawn_from: Neurospora crassa
    pool: taxon:367110
    method: census
    detail: all 16 GO/function records, plus 4 localization cases
  - id: POMBE20
    label: POMBE20
    drawn_from: S. pombe, not on the list
    pool: pombe_export
    method: off-list
    detail: Swiss-Prot, from the pre-release export
    # Eligible here is the 20 GO/function-bearing export entries, so 20 -> 20 holds by construction.
  - id: POMBE_REMAINING8
    label: POMBE remaining 8
    drawn_from: S. pombe, not on the list
    pool: pombe_export
    method: off-list
    detail: the other 8 entries (names and locations only); queued
    eligible: complement
  - id: MOD_EVOLUTION20
    label: MOD_EVOLUTION20
    drawn_from: 7 model species
    pool: mod_species
    method: hand-picked
    detail: "evolutionary questions: substrate, catalysis, complexes"
  - id: DONKEY40
    label: DONKEY40
    drawn_from: Equus asinus
    pool: taxon:9793
    method: ranked rule
    detail: most GO terms; needs TreeGrafter and horse IBA
    post_snapshot: true
not_drawn_from_release:
  - HORSE40_HUMAN_PAIR