I am trying to use SCVI. I normally use scanpy and SCVI is quite new to me. When I read the SCVI tutorial, I noted that for high var calculation, SCVI use totally raw data. My adata is soupX corrected data (not completely raw). In that case, how can I use SCVI? Can I input soupX corrected count matrix data to SCVI in any way? This is both for batch correction and label transferring (also for cite-seq data analysis). Would appreciate your advice on that as I don’t have completely raw data.
I think the definition of raw here is ‘unnormalized’, which is indeed the case if you ask SoupX to give an integer matrix. I would say it’s okay for scRNA-Seq data, though I am not sure about CITE-Seq.
Agree, though I’m not an expert on SoupX. And by CITE-seq, for the RNA part it would be fine, for protein I don’t think it makes sense, but I stand to be corrected.
You would treat the soupx as raw in the scanpy hvg function, there’s a flag if it complains about integers.