[BioC] DESeq for an mRNA-seq time course

[Simon Anders]

Hi Elena

With this here

    fit1 <- fitNbinomGLMs(cds, count ~ treat.type + time.hr)
    fit0 <- fitNbinomGLMs(cds, count ~ treat.type)

you are testing against the null hypothesis that a gene's expression is 
independent of the time of the sample, i.e., you will get a hit whenever 
you can reject the claim that the gene has the same expression in all 
five time points. This is a pretty narrow null hypothesis. It seems 
quite plausible to me that around 50% of the genes change noticeable 
over time.

Note also that treatment is a factor in both your full model and in your 
null model, so you are not testing whether the treatment has an effect. 
I suspect that this might not be what you want.

In a simple comparison of treatment vs control the question one wants to 
test is usually fairly obvious: does the treatment influence the 
expression or not. For time course data one usually does not have such 
an obvious yes-or-no question, and hence, you may need to tell us more 
about what exactly you want to test before we can advise you.

   Simon