I did a lot of validation work to show that the approach works when using pseudobulked data. I even implemented bootstrapping to get estimates of confidence. This all got bundled into an R package which I tried to make pretty easy to use. This is an osteo patient sample.

Just a preview of something I should have out soon. I'm putting together a paper using genetic variants in single cell RNAseq data to distinguish tumor from normal cells. I show why calling variants on single cells is a bad idea (tm). Relaxing normal cutoffs leads to inaccurate variant calls.

Day 1 down! Four to go

That's a very cool demo. Thanks for writing it up. I tried it and asked it to summarize some limitations of the package and got this, which I though was amusing since the first commit for the repo was 2023...