I am using edgeR to analyze RNASeq data and I wonder about how the interactions between two factors is dealt with, especially when the normalized cpm values tend to 0 on one of the factors.
I have the factors “time” (T0/T30) and “genotype” (A,B,C). In order to build my contrasts I created a dummy factor comprising all possible combinations among levels of those two factors.
Please find attached a table with 4 genes on which different contrasts (in columns) have been tested. The coding on the table tells us whether the given comparison is significant (i.e. “1” when BH<0.05) or not (“0”).
Please find also some plots where the normalized cpms for those genes are shown. The lines join the genotype-wise mean cpm values on each time point.
For those four genes, in my view, time does not play the same role on all genotypes, and yet the table tells me that edgeR does not find the interaction between factors significant. Is this because the cpms for those genes are close to 0 when T0? If this is the case, why is so? And if that is not the reason, what could it be? I guess Gene3 is worth mentioning cause cpms on genotype A are different from zero and yet no interaction is significant between time and genotype.
Thanks for your help on this.