EPIC: Write Makefile to drive RDF-OWL generation for KG construction from CxG corpus

[dosumis]

Pipeline will generate rdf-owl files by running Pandasaurus_CxG across CellXGene datasets accessed via CxG census.
Configuration of labelsets will come from:

• Mary's curation of integration datasets for CAP
• Bradley's curation of CxG datasets: mapping file to be added to this repo
• Future curation effort on CxG-CL project.

Some work to extend pandsaurus_cxg may be needed

• to pull in additional metadata.
• to add support for lookup via census

TBD: should we be storing cell ID lists?

[dosumis]

MVP:

1. Write Makefile that:

   • reads CSV input - detailing datasets and author_category fields (CSV to be provided by @dosumis)
   • Pulls Anndata file from Census
   • Runs Pandasaurus to generate RDF

2. Run OBASK,

   • importing all output RDF files from step 1.
   • Ontologies: CL; GO; Uberon

Next steps:

1. Extend Pandasaurus to pull more ontology metadata from CxG AnnData & to include dataset metadata in RDF output.
   • Add dataset node. Populate with metadata in uns + UUID of dataset. <-- DO THIS FIRST.
2. Write Python module to pull GO annotations from the QuickGO API for the GO terms with direct links to CL terms in graph. @dosumis to provide spec.
3. Write import pipeline to pull in NS-Forest markers from Richard Scheuermann' analysis (markers should be looked up in anndata var).

Challenges:

Reconciling/linking gene & protein IDs. These may come in as PRO, uniprot, ensembl + we need to record species. Decision to make: aim for unified gene nodes that collect IDs? (ID label pairs?)