It is actually simple. If you have 10 vs 1 then do 10 vs 1 with
DESeq2 or alternatives such as
edgeR, but be aware that the lack of replication in the control will inevitably lead to more false results compared to a well-replicated design. You cannot estimate the variability in the controls and therefore results are not reliable. Consider results as exploratory and try to validate interesting candiates by independent experiments, don't make strong claims on underpowered experiments. Nothing more you can do here if you ask me. Not sure about DESeq2 (maybe the developer has implemented an option to forbid these kinds of testings) but edgeR will technically allow this comparison, if you are willing to accept the mentioned shortcomings.
modified 12 weeks ago
12 weeks ago by
ATpoint ♦ 44k