Exact test vs glmQLFit for comparisons between two groups

Question

[catscampos]

Hello everyone,

I'm working with EdgeR, and I have two RNA-seq groups to compare, with only two samples each. I've used exact test and glmQLFit test, since glmQLFit can be useful when the number of replicates is low. I get very distinct number of DGEs using the two approaches (much more in the exact test), although with the same trend (much more up-regulated in the same group). Do you know which one is better to use?

Thanks in advance!

Catarina

Answer 1

[Kristoffer Vitting-Seerup]

Without a doubt glmQLFit!

1. glmQLFit is the 3rd generation of tests in edgeR - and the edgeR guys don't update unless it is improvements. EdgeR was classically views a beeing to liberal (not controling it's false discovery rate (FDR), so even though you use a cutoff of say 0.05 the actual FDR is way higher). This is fixed with glmQLFit.
2. That is the suggest way to analyse data in the vignette :-)

[catscampos]

Thanks a lot! Very helpful!

Catarina