Question about how to interpret DEXSeq data with a multi-factor experimental design

[mlloyd]

I have successfully run DEXSeq (version 1.16.7) in R version 3.2.3. I believe the analysis ran smoothly without any problems but I am having trouble interpreting the results.

I have a multi-factor experimental design:
Day 1/Low CO2
Day 1/High CO2
Day 7/Low CO2
Day 7/High CO2

I checked for the effect of CO2 with the following formulas:
~ sample + exon
vs.
~ sample + exon + CO2:exon
This resulted in 134149 insignificant hits and 3 significant hits

I then checked for the effect of CO2 when accounting for Day with the following:
reduced model = ~ sample + exon + Day:exon
full model = ~ sample + exon + Day:exon + condition:exon
This resulted in the same results as above

I then did 4 separate tests subsetting the data:
Effect of Day in only low CO2 data: 4,821 significant hits
Effect of Day in only high CO2 data: 4,258 significant hits
Effect of CO2 only in Day 1: no significant hits
Effect of CO2 only in Day 7: no significant hits

The large effect of Day is expected because we were sequencing early embryos that were developing very quickly. However, you can see the number of significant exons is not the same in the two tests for the effect of Day. Furthermore, when I compare the lists of the 4,821 significant hits to the list of the 4,258 significant hits, I find a lot of hits unique to each test.

This leads me to believe there might be an interaction between CO2 and Day but I cannot figure out how to test for it. My question is: How do I test for the interacting effect between two factors in DEXSeq?

Thanks!