Hi,

I've been asked to do a Differentially expressed genes analysis using RNA-Seq from TCGA database. I've been using DESeq2 for all my previous analysis on cell lines.

I'm having an issue with the control samples selection in TCGA. I've seen that DESeq2() should not be used with a mix of paired and unpaired tumors and controls dataset and that limma-voom duplicateCorrelation() should be used instead.

If have a dataset that has, let's say 50 tumors, with condition A or B, 5 paired controls and 100 unpaired controls, can I just remove the paired controls and use DESeq2 then? I have controls to spare here and would like to keep the same software because I'd like to use the same methods I've used for the cells.

Have a nice day!

Yes I think dropping the 5 paired in this context isn’t a loss of much information.