I am working with DESeq2 to perform a differential expression analysis between different treatments.I have 4 conditions and 4 four biological replicates for each conditions.
I’ve performed differential expression analysis using DESeq2 building a dds model considering all conditions together and subsequently I have extracted the pairwise comparisons between treatments.
In addition I have created a specific dds model, subsetting a DESeqDataSet to only samples from those the group that I want to compare and then I ran DESeq on this subset.
After I ran the two analysis approaches, we have observed that we identify an higher number of DE genes using the DESeqDataSet that includes the 4 conditions (174 DE genes) than when we perform the analysis considering a DESeqDataSet that includes only samples of the two conditions that we compare (103 DE genes). I was wondering why performing the analysis with a model including all samples is more sensitive.