Scirpy side initiation

[vd-dragan21]

Understanding ir.datasets.vdjdb() and Scirpy's Data Needs

• Go through scirpy documentation
• Go through Jupyter Notebook from Francesca
• Have a look at GitHub issues:
  Hackathon idea for collab issue ‘23: Hackathon: add generic_ir_from_biocypher() function to ingest TCR data scverse/scirpy#404 (comment)
  Scirpy needs more databases issue ‘21-’24: Add more epitope databases to datasets scverse/scirpy#308
• Read the preprint on TCR specificity methods: https://www.biorxiv.org/content/10.1101/2024.10.26.620398v1
• Meeting with Gregor
• Document requirements from scirpy side

[vd-dragan21]

Gregor call sum up:

1. There are a lot of databases of interest, that ideally all should be merged but the most important are VDJDB and IEDB:
   AIT, huARdb, McPAS-TCR, PIRD, Adaptive Biotechnology, SARS-CoV-2 resources, VDJDB, and IEDB: https://github.com/scverse/scirpy/issues/308

2. "Real data curation effort" from the following issue is about a database bug and can be ignored now: https://github.com/scverse/scirpy/issues/404#issuecomment-1888578330

3. Current approach: download the database and save in the cache

4. Wanted approach: accessing a non-static database (KG?) made of several merged databases

5. Desired output: python dictionary/json file, Gregor will convert it to the AnnData himself

6. There is a banch of Metadata but the most important ones for now are: species and genes for filtering: https://docs.airr-community.org/en/stable/datarep/rearrangements.html#fields

7. The preprint paper recommended by Gregor (https://www.biorxiv.org/content/10.1101/2024.10.26.620398v1) is not very useful. They merged 3 databases statically in a table. Yes, we are interested in these dbs but not the static ones:
   "The dataset used in this analysis is based on a combination of three publically available databases:
   Immune Epitope DataBase (IEDB)[16], VDJ database (VDJdb)[17] and the manually curated
   catalogue of pathology associated T-cell receptor sequences (McPAS-TCR)[18], with the data being
   collected as of March 2023."

...Mb have a look at the code for merging/scoring later: https://github.com/i3-unit/TCR_Unsupervised_Benchmark

8. Gregor is not actively working on the project but works on scirpy in his free time, so that is why there won't be any regular updates/meetings with him but he is always happy to help with certain questions

Sum sum up of the project goal:
Public BioCypher API that one can query (Realistic goal: 2 databases: VDJDB and IEDB and 2 metadata objects: species and genes)