Combined p-values for DE genes from RNA-seq meta-analysis

Question

[ks]

Hello, I have raw p-values from edgeR for two comparisons of interest: control vs genotype1 and control vs genotype2. I'd like to see if there is an aggregate effect of genotype on DE.

The Internet has a number of answers on how to do a Fisher's combined test, and some resources on how to do this in R (such as the metaRNASeq package). But edgeR (as well as DESeq2) give two-sided p-values, while the R functions for combined tests requires a one-sided p-value. Does anyone know how to proceed? I'm also not sure if this is a valid way to get an aggregate result, since these two tests have the same control samples and are not independent. Does anyone have thoughts on whether it would be better to run the analysis in edgeR again, treating genotype1 and genotype2 as one group?

[chris86]

You haven't explained which studies are separate? Are control vs genotype1 and control vs genotype2 two completely separate experiments with new controls and new genotype samples?

[ks]

No, they are from the same experiment -- the samples sequenced were control, genotype1, and genotype2.

Answer 1

[chris86]

You can use limma for RNA-seq analysis now it has been updated. You can see in Aarons answer on this page he describes how to do contrasts for a A+B versus C comparison which is what you seem to want to do. I think this is what you are looking for - at a guess.

con.2 <- makeContrasts(A - (B+C)/2, levels=design)

https://support.bioconductor.org/p/69764/