Ontologies

The biomedical source ontologies behind the three pairs, with where to obtain each and under which licence. Nothing is re-hosted here.

OAEI Bio-ML matches whole biomedical ontologies. To keep licensing clean, the track does not re-host any source ontology: each is obtained from its original publisher under that publisher’s licence. This page lists the sources, their licences, and which pair needs which. The reference alignments themselves are distributed with the task data.

Which ontologies each pair needs

PairOntologies to obtainAlso used in reference construction
NCIT–DOIDNCIT, DOIDMONDO, UMLS
SNOMED–FMASNOMED CT, FMAUMLS
SNOMED–NCITSNOMED CT, NCITUMLS

NCIT, DOID, SNOMED CT and FMA are the four ontologies matched across the three pairs. MONDO and the UMLS Metathesaurus underlie the reference construction (the gold references are grounded in UMLS/Mondo).

Sources and licences

Three sources are openly downloadable; three are access-controlled and require a licence/account.

Openly downloadable

Access-controlled (licence / account required)

  • SNOMED CT (clinical terminology; licence: SNOMED CT Affiliate Licence) A comprehensive clinical healthcare terminology from SNOMED International. Access requires a SNOMED CT Affiliate Licence — free of charge in UMLS-member territories via the UMLS Terminology Services (UTS), or otherwise through SNOMED International.

  • UMLS Metathesaurus (metathesaurus; licence: UMLS Metathesaurus Licence via UTS) The Unified Medical Language System integrates many biomedical vocabularies; it underlies the reference construction for every pair. Requires a free UTS account: uts.nlm.nih.gov — see the UMLS overview.

  • FMA — Foundational Model of Anatomy (anatomy; per its own licence) A reference ontology of human anatomy from the Structural Informatics Group, University of Washington. Obtain per its licence from the FMA project page or BioPortal.

Notes

  • Use the ontology versions pinned by the 2026 edition where indicated on the dataset card; matching against a different release may shift entity IRIs.
  • Track 1 correspondences use full OWL IRIs; Track 2 uses CURIEs (e.g. NCIT:C101044). This is intentional — see evaluation metrics.
  • Nothing on this page is re-distributed by OAEI Bio-ML; all links point to the original publishers.