PHENIO / uPheno ingestion
VariantHound uses PHENIO as the cross-species phenotype backbone for deterministic semantic similarity. PHENIO is an application ontology that includes uPheno and multiple species phenotype ontologies; VariantHound does not copy the complete integrated graph into its public dataset.
Pinned inputs
data/manifests/phenio-2026-09-01.json pins PHENIO release v2026-09-01 by tag, source commit, immutable GitHub asset IDs, byte counts, and SHA-256 digests. The transform consumes:
phenio-relation-graph.gz, the release's materialized relation graph; andphenio-upstream-versions.tsv, which records the integrated ontology versions.
The relation graph was generated upstream with equivalence-as-subclass, subclass output, and reflexive subclass relations. VariantHound preserves that meaning as an entailed-reflexive-subclass-closure rather than treating the rows as asserted direct edges.
Compact output
pipelines/ingest_phenio.py scans all 82,128,289 relations but retains only the 3,972,487 subclass-closure rows needed for human/mouse cross-species comparison:
| Subject | Retained ancestors |
|---|---|
HP |
HP, UPHENO |
MP |
MP, UPHENO |
UPHENO |
UPHENO |
The output phenotype_closure.tsv.gz contains phenotype_id and ancestor_id. It includes one reflexive row for each retained phenotype, enabling exact self-matches. upstream_versions.tsv keeps the release-level provenance for all integrated sources. Labels and asserted local hierarchies remain in the separately pinned HPO and MGI snapshots.
Relations to anatomy, qualities, diseases, genes, and phenotype ontologies not used by VariantHound's first ranking path are excluded. This keeps the processed package compact without replacing PHENIO's inferred cross-species bridge with a hand-authored mapping.
Evaluation scope
This is a current release backbone, not a historical snapshot. It may be used for current retrieval and discovery simulation when case-specific evidence is masked before feature construction, but it cannot support temporal evaluation of older cases.
Reproduction
Place the two pinned assets in dist/phenio/raw/, then run:
python3 pipelines/ingest_phenio.py
Use --download to fetch the immutable release assets. Every source checksum, relation count, prefix-pair count, reflexive-term count, and output checksum is verified before the snapshot is accepted.