I have three groups each have 4 replicates. The conditions for each group were same before preparing library and then sequencing. Now I have to analyse the read counts to find DE genes between Group1 vs Group2, Group2 vs Group3 and Group1 vs Group3.

But I do not know whether to combine all the groups in one file then perform RUVs normalization and find DE genes in each case OR make a separate file for each case, then perform RUVs normalization separately on each file and then find the DE genes in each case.

Thanks!

Hi,

it is definitely best to combine all the data into one matrix, normalize, and test for differential expression in one go. This will allow you to estimate the variance across all of your samples, which should give you more power.

Note that since you are interested in the pairwise differences, you will need to use contrasts. Both the edgeR and the limma user guides have good examples on how to use contrasts to test for the pairwise differences (See for instance section 3.2.3 of the edgeR user's guide).