Hi,

I am analyzing a time course experiment and I am following the example in the DESeq2 tutorial as my experimental design is very similar in that I have two strains sampled at several time points. I would like to identify genes whose expression is different between the two strains at any time point and whose expression displays different trends in the two strains across time. At first, I tried doing the analysis the way it is suggested in the tutorial:

ddsTC <- DESeqDataSet(fission, ~ strain + minute + strain:minute)
ddsTC <- DESeq(ddsTC, test="LRT", reduced = ~ strain + minute)

and I ended up with a relatively small number of genes.

Then I tried removing strain from the reduced model as I wanted to include genes that are DE between strains regardless of the time point:

ddsTC <- DESeqDataSet(fission, ~ strain + minute + strain:minute)
ddsTC <- DESeq(ddsTC, test="LRT", reduced = ~ minute)

and I ended up with a much larger list of genes. Interestingly, in this way I also captured many additional genes which display different expression trends over time in the two strains. Is there any reason not to do what I did here?

Also, is there any advantage in doing what I did over conducting simple pairwise comparisons between strains at each time point and concatenating the lists of DE genes?

Thanks!

The choice of the null hypothesis is up to you, and will determine the list of genes. In the second one, the null hypothesis is that strain and its interaction with time have no effect on gene expression: that gene expression only changes over time, and that the two strains have identical expression level at the reference time point.

Concatenating pairwise comparisons per time point should give you a similar list, but will have less power.