Hi all, we performed a RNA seq experiment to compare the same cell line treated with drug1 and drug2. Now we decided to compare these expression profiles to expression profile from untreated cell line but I am confused about the right approach. I would proceed by creating two different count matrices from salmon results, then merge them and import the final one in deseq2 (DESeqDataSetFromMatrix) considering to include in the final design formula the experiments from where the samples come from (design = ~condition + experiment ). Has it sense?

I tried to import with salmon all the files (those from drug1, drug2 and untreated) with tximport and to use a design formula considering the experiments (design = ~condition + experiment ). The point is that comparing drug1 vs drug2 return different results to those obtained by following the same approach but importing only salmon files from drug1 and drug2.

If you want to estimate two different effects, you would use ~experiment + experiment:condition, which will control for experiment differences at baseline and estimate two effects, see resultsNames(dds) afterward.