Hi,

I have the following RNA-seq experiment:

control (right tissue) vs control (left tissue)

case (right tissue) vs case (left tissue)

I would like to find the DE genes present in the first comparison + DE genes second comparison but discarding those DE genes that are significant in both comparisons with the same fold change direction. The way I proceed is to perform DE analysis for each comparison separately and then removing the overlapping ones manually. But I would like to know if there is a more direct way performing only one test and extracting the required contrast.

So far I tried this with DESeq2:

design~group

res <- results(dds, contrast=c(0,-1,1,1,-1))

resultsNames(dds)
"INTERSECT" "GROUPctl_LEFT" "GROUPctl_RIGHT" "GROUPcase_LEFT" "GROUPcase_RIGHT"

But the contrast is not the desired one.

Anyone could help me?

There is a not a contrast that can give you the genes which are DE in the first comparison but not those which are also DE in the second with the same sign of log fold change. That's just not a linear combination of the coefficients in the model (that's all a contrast is).

You could combine p-values from two tests: the first, a contrast of the first two groups, and the second, you want to require that the difference in the L vs R across group is in the opposite direction of the LFC for control, correct?