Dear all,

I am working with 19 RNAseq sample, 10 of which have 2 technical replicates (same individuals in a population).

I am trying to correct for a clear batch effect using the RUVs approach as described in http://nar.oxfordjournals.org/content/early/2015/07/21/nar.gkv736.full and in the tutorial https://github.com/drisso/peixoto2015_tutorial/blob/master/Peixoto_Additional_File_1.Rmd .

How should I combine these technical replicates after normalisation to avoid pseudo-replication when I do the differential expression part?

Do the factors of unwanted variance, found after RUVs, already take into account the unbalance in the replication design? If not, is there a way to include this info in the glm design given to EdgeR (or DEseq) when we do the DE analysis.

Thank you for your help

Thomas