I have a question regarding different results when using contrast in DESeq2 vs when I do the analysis separately.
I have 3 groups : A : condition A; n = 25 B : condition B; n = 25 C : controls; n = 350
I first looked at differently expressed genes between A and C, and B and C separately. I performed the analysis and then retrieve the separate results for each analysis using contrast.
However, if I use contrast to compare A and B, (when C is considered in the dispersion analysis) vs performing an isolated analysis for A and B, the results are not the same. (no FDR genes with the first approach vs plenty with the second)
I am wondering if the number of controls "masks" the difference I could identify between A and B when it is considered in the dispersion.
In this case, I am wondering if performing an independent analysis for A an B is better.
Many thanks !