Hi All, I am interested in identifying bacterial genes that are differentially expressed in vivo in comparison to the in vitro bacterial culture. However, I encountered a challenge regarding the sequencing depth of my samples. For in vivo bacterial RNA samples(total 8 samples) the sequencing depth was low in the range of 250K-1.3 million reads due to the presence of a large amount of host mRNA. For in vitro bacterial cultures (3 samples) the total reads were in the range of 70-100 million. My question is can DESeq2 handle such huge variations in sequencing depth among the samples? If not, can I down-sample the in vitro bacterial reads perhaps around 10-20 million may be using seqtk tool? I greatly appreciate any inputs you can provide on this matter. Thanks, Gopi Viswanathan

In the 2014 paper we note there is an issue if depth is confounded across condition.

One solution to this design issue is that you could only look at features that have a minimal number of counts in the under-sequenced samples, e.g.:

keep <- rowSums(counts(dds[, dds$condition == "X"]) >= 10) == sum(dds$condition == "X")
dds <- dds[keep,]