I have data for 26 different tissues and I want to perform exhaustive
pairwise comparisons between them. I want to find normalized expression
levels for the samples and differentially expressed genes between tissues.
For both I intend to use DESeq2.
From my reading and testing, I believe that the samples have two
o the replicates of a two tissues show high 'within-group variability'.
The manual explains to me that this can bias the dispersion calculations.
o the sample data have a wide range of coverage: when I run DESeq2 on all
of the samples at once and calculate the size factors, I find a range
of .1 to 41.
My questions are
o is it reasonable to use DESeq2 expression level values rather
than using read counts to normalize samples?
o is there an objective way to decide whether to run the differential
expression analysis 'all together or split into pairs of groups?'
(I have not found satisfying answers to these questions in my web searches
and manual reading.)
As always, I am grateful for assistance and consideration.