quasi-hierarchical cell typing

[patrickjdanaher]

The basic idea: to differentiate between closely-related cell types, we should re-run insitutype using only relevant genes, not the whole panel.

Rough workflow:

1. run insitutype under whatever workflow you want - sup, semi, unsup
2. automatically identify supersets of related (i.e. occasionally confused) cell types
3. for each of the supersets, identify useful genes: decent expression, no contamination risk, maybe also hvg
4. then run supervised within each superset, using the profiles from the original round, and the selected subset of genes

The final output won't have logliks on the same scale, but the posterior probabilities will still make sense, even if they ignore the (usually tiny) probabilities of belonging to a cell type outside the superset.

Steps 2-4 could be called as a single function after any basic insitutype run.

[github-actions[bot]]

Thank you for contacting us about our tools! To receive assistance, kindly email support@nanostring.com with detailed information about your issue. If applicable, attach a screenshot of any encountered errors and include a copy of the modified script in Notepad. Our customer support team will help facilitate a review and resolution of the issue.

Thank you for choosing NanoString, NanoString Dev Team