Why is zscore calculated for unannotated barcodes?

[johnchamberlin]

I'm trying to use readzs and it is currently generating no significant windows. While investigating, I noticed that most of the zscores correspond to barcodes that are not annotated with any 'ontology', even in the small test data example. Why is this the case? Presumably these correspond mostly to empty droplets or noise.

Here you can see that 58% of barcodes in the test dataset have a 'tissue' annotation.
> data.table::fread("small_test_data_chr21.zscore") %>% summarise(mean(!is.na(tissue))) mean(!is.na(tissue)) 1 0.5862645

For context I am trying to do a comparison between only two cell types within a larger dataset, so in my case the vast majority of cells are not annotated - I only provide metadata for the barcodes I care about. Do the zscores for unannotated barcodes affect the calculation of significant medians in any way?

If I run readzs on the full dataset, there are many significant windows, but I am trying to do pairwise comparisons for each cell type between two experiments, so the 'one vs all' comparison is irrelevant.

thanks,