BG taxonomy hacking experimental_notebooks

Aims:

• Check CL term curation up to date

• Test flattening human BG tax + h5ad - for CAP Notes:

  1. It seems that everything is flattened. We should review this, it conflicts with option 2 here: cellannotation/cas-tools#111 Decision - we will stick with the status quo for now, but need to review. The flattened files are ugly, but we can offer unflattening. TODO - check what happens when we flatten annotation transfer.
  2. We are getting some warnings about hashes.
  3. TDT issue:
     • Hierarchy merges down where two adjacent levels have the same name. Needs to be fixed. <-- v.high priority. Action item: Fix TDT issue and then share flattened human BG taxonomy with CAP. Action term: Review roundtripping - should we switch to mixed strategy?

• Add cell set level AT to Macaque BG tax

  • Clusters get the most specific AT (homology subclass)
  • Everything else gets most specific matching block - if one exists
    • e.g. for
    • 
    • cluster:
      • 41_IN:Lamp5 Lhx6 gaba
      • 42_IN:Lamp5 Lhx6 gaba
      • 43_IN:Lamp5 gaba
    • primate subclass:
      • CCK-FBXL7: CTX-MGE GABA
      • CCK-VIP-TAC3:vip Gaba
      • CHAT:PAL-STR Gaba-Chol
      • D1-ICj:OT D3 Folh1 Gaba
    • primate-neighbourhood:
      • Astroyctes: Astro-Epen
      • CCK:CTX-MGE GABA
      • TAC3:CTX-CGE GABA
      • cholinergic: PAL-STR Gaba-Chol

  If there are mutiple matches on the same level, we add all, again matching most specfic block

  

  41_In : 287 + 288

  CCK-FBKXL7 : 287 + 288

  etc

  Consider adding a gropuing mechanism in future

  • Current schema:

    • transferred_annotations (list)

      • transferred_cell_label (string): Transferred cell label.
      • source_taxonomy (string): PURL of source taxonomy.
      • source_node_accession (string): accession of node that label was transferred from.
      • algorithm_name (string): .
      • comment (string): Free text comment on annotation transfer.

    • Notes - we need a new field for labelset

    • We need confirmation of algorithm and some text for the coment from Nelson

  • Code with new CAS-LinkML lib.

    • LinkML misfeature means that author annotation fields get stored as JSON_to_string which is a not a JSON string. Solution for now - we convert to JSON string before loading into CAS-LinkML instance. Convert back at end.

• (Test) Merge Macaque BG tax with new h5ad