Closed brianraymor closed 6 months ago
Per the June 13 tissue mappings - there can be only one call with @BAevermann @jahilton @signechambers1 @pablo-gar :
@jahilton proposed an ordered list model -
[
most narrow UBERON term which conceptually maps to curated tissue general / organ,
... ,
most broad UBERON term which conceptually maps to curated system
]
Cell Census and Gene Expression will prefer the most narrow term. Discover UX Filter will prefer both most narrow and broad.
The curated lists will be documented in the schema and updated on a regular cadence.
Next steps:
Per cell census schema triage, closing schema mappings in favor of returning mappings in an ontology service.
See Integrate all required ontology processing into ontology preparation
Re-opening per conversation on #cell-science-census.
Not required for the dataset schema.
Context
See single-cell-census.
Definition: _high-level mapping of a tissue, e.g. “Heart” is the tissuegeneral of “Heart left ventricle”
All high-level mappings (1-1 or tree) of tissue MUST be defined in the dataset schema. When a dataset is uploaded, cellxgene Data Portal MUST automatically add the relevant mapping. The results will be consumed by all cellxgene experiences for consistency.
Two tissue mappings are implemented in CELLxGENE Discover:
tissue_general
andtissue_general_ontology_term_id
are defined in the census schema.Research
The Tissue mappings and the cellxgene-schema CLI all calculate ancestors and need a common approach. Also see:
There are open questions in Proposal: Standardization of ontology-derived process and assets of CELLxGENE Discover about whether such information is required to be embedded in datasets rather than be returned by an API.