Hi,

I'm doing a DE analysis with multiple conditions and I wondered what the right way is of doing so. For each comparison I'd only like to use a subset of the data. I wondered if it's possible to make one DESeq2 object and subset it for each comparison, for example like this:

dds <- DESeqDataSetFromMatrix(countData = cntmtx,
                              colData = colData,
                              design = ~ condition_A + condition_B)
dds <- DESeq(dds)
dds_sub_x  <- dds[,dds$condition_A == "x"]
res_sub_x <- results(dds_sub_x, contrast=c("condition_B","a","b"))

If condition_A has the levels of "x" and "y" and condition_B has the levels of "a" and "b".

Could someone tell me if it's alright to go about it like this?

You could, but it's not necessary, and it's probably not optimal. You'd probably rather have all the samples for library normalization and dispersion estimates.

If you only want to compare a subset of the samples to another subset, make a new column in Coldata which concatenates the data from conditions A and B, (so x_a and x_b), use that column as your design