Hello everyone,

I have a dataset of RNASeq counts for several cell types. For each cell type, I am running a DESeq model for the specific subset. Should I run estimateSizeFactors for each subset or use the global one?

Example:

dds<-DESeq(dds) 
dds_sub<-dds[dds$condition!="CellA",]
dds_sub<-DESeq(dds_sub)

or...

dds<-DESeq(dds) 
dds_sub<-dds[dds$condition!="CellA",]
dds_sub<-estimateSizeFactors(dds_sub)
dds_sub<-DESeq(dds_sub)

Thanks in advance.

I usually use the same size factors estimated on one dataset for all comparisons, likewise the dispersion estimates. See the FAQ in the vignette on this question.