Prediction using posterior sampling

dissatisfaction-ai / scHierarchy

A toolking for cell type hierarchies: marker selection & consistent classification

Apache License 2.0

20 stars 1 forks source link

Prediction using posterior sampling #7

Closed vitkl closed 2 years ago

vitkl commented 2 years ago

According to Adam Gayoso, this code is not ideal because it ignores adata object validation that they worked a lot to set up. However, it seems to me that their approach doesn't work for us because we use slots with custom names.

TODO:

[x] Add test for prediction task.
[ ] Pick the most likely class for each cell at each level and put that into adata.obs column for corresponding level_keys (since we have to create those columns we might as well fill them).

vitkl commented 2 years ago

Actually, I found the solution - the object with cells for which we want to predict class needs to be not registered AND have the same input columns.