Dear Michael,

I have a full factorial  experimental design with 2 factors that have 2 levels each, a total of 4 combinations.

I am interested in the interaction between those two factor and that is the reason why I used the following command:

dds <- DESeqDataSetFromMatrix(countData= reads,colData = Design, design = ~ Factor1 + Factor2 + Factor1: Factor2).

Afterward I got a list of contigs that are significantly interacting between the two factors.

My questions are:

1) is there a way to test within the same model the independent effect of each factor or shall I rerun another model without the interaction?

2) with this complete model I can get lists of differentially expressed contigs from the comparison between two level of one factor within each level of the other factor separately and reciprocally (total of 4 comparisons). This looks like a post hoc multiple comparison test (which is allowed only if the interaction is significant). So shall I substract from those lists of contigs the ones that are absent from the list of interacting contigs or not?

Than you for your help.

Matteo

(1) Depending on what you mean by independent, then no. The interaction model means that the fitted effects for factors 1 and 2 take into account their interaction (you obtain the effects for the reference level of the other factor). See the interactions diagram in the DESeq2 vignette. 

(2) Different people approach this different ways, but my preference would be to present results from three tables separately, with multiple correction within each table. The three tables of interest in such a setup in my opinion are, e.g. the condition effect for group 1, for group 2, and the interaction (whether the condition effect is different). Since I know ahead of time that I will present all three tables, I'm not doing any changes to the analysis based on what results I obtain, and I correct for multiple testing within each table, so the reported FDRs can be interpreted appropriately.