if I want to find genes showing any sign of interaction between time and genotype, could I simply do this: glmQLFTest(fit, coef="interaction column indices"), as said in "edgeRUsersGuide.pdf"'s "3.3.4 Interaction at any time"?
I have two genotypes and six time points. So there are five interaction columns in the design matrix. But when I looped over all combinations of the interaction column indices, i.e. five 1-index combinations plus ten 2-indices combinations plus ... and finally plus one 5-indices combination, and collect genes with FDR < 0.05 in each loop, I found that the overall unique genes with FDR < 0.05 from the looping are more abundant than that of the one 5-indices combination.
So I am confused which way is better doing my job.
Thanks a lot.