Frequently asked questions
Short answers to common pre-launch questions. Details that aren't final are flagged provisional and linked to the page that owns them.
Is the dataset available yet?
Not yet, but it will be available shortly. The package — the graph, candidate sets, and public labels — is described on the Data page; the hosting platform and release timing are still being finalised. Provisional
Is OWL expertise required?
For the main track, no — it is framed as typed link prediction over a knowledge graph, so no OWL or description-logic background is presumed, and it is meant to be broadly accessible. The complex track is different: it is OWL class-expression generation and does require that background.
Can external resources be used?
The main track is open — use any systems and any data to produce your ranking. The draft policy asks that external resources and pretrained models be public, cited, and declared. See Participate for the rules.
How is the hidden ground truth protected?
The test split releases candidate sets only — the gold targets are withheld. Final scoring runs server-side against those held-out answers, so a submission is judged on labels it never sees.
What is the complex track?
A separate track on generating OWL logical definitions rather than ranking candidates. Systems reconstruct the owl:equivalentClass definitions of the Human (HP), Mammalian (MP), and Worm (WBP) phenotype ontologies in terms of background OBO Foundry ontologies (anatomy, cell types, chemistry, processes, qualities). See the complex-track guide (more information coming soon).