Merging multiple scRNAseq samples: Normalize first and then merge or merge and normalize all together?

Question

[Roman Hillje]

Hi,

I was hoping you guys could give me some advice on single-cell RNAseq analysis. I have four samples that belong to the same experiment. They were generated with the 10X Chromium (DropSeq) technology but I'd like to do the analysis independent from their software. Currently, this workflow is based on the scran and scater packages you can find on Bioconductor.

My direct question is: When merging the cells from different samples into one "project" (which is necessary to compare the populations), do I first prepare each sample (filter cells, normalize molecule counts) and then merge them, or do I first merge the raw molecule counts and then filter + normalize all together?

Is my question clear?

Thanks in advance!