I am using DESeq2 to find differentially expressed genes between conditions control/treated. The PCA using vst counts shows family-based clustering.
I have conjured up a few different GLM models as below along with the number of differentially expressed genes detected:
Model Num of DEGs
(using principal components from pca above)
(sva was used to generate one surrogate variable)
A venn diagram of the DEGs from the 5 models looks like:
Is there some way to objectively evaluate (quantify) which model is better?
This question is related to this.