[BioC] DEXSeq: Problems trying to reproduce the example of the vignette

[Simone]

Dear Wolfgang,

thank you very much for your reply.

> plotDispEsts
> as a workaround, please run the R code in the vignette from the
> .R file provided with the package

I missed this! Now it works fine.

> I am not sure I understand your problem with 'plotMA'; what do you get from
> table(res$padjust, useNA="always")

Indeed, there are NA-values in the table. But when I remove all NAs of
the data.frame the plot still behaves different when adding the color
parameter. Different data points than without setting the parameter
seem to get highlighted and additionally some data points disappear. I
still do not really understand what's going on here and how I should
effectively change the coloring regarding the p-value threshold I want
to use in my analysis without causing such "side effects". Any
suggestion? Sorry for the confusion, but I would really like to
understand this.

> As for your third question, isn't this solved by just adding another
> covariate for 'patient' which will absorb the patient effects?

Well, yes, I also thought that I'd do it like this, but just wanted to
be sure that I am not missing any additional parameter I would have to
set explicitely for paired tests or something. So simply adjusting the
code to the following should be the right way to account for paired
patient samples, shouldn't it?

formuladispersion <- count ~ sample + (condition + patient) * exon
myExons <- estimateDispersions(myExons, formula=formuladispersion)
myExons <- fitDispersionFunction(myExons)

formula0 <- count ~ sample + patient * exon + condition
formula1 <- count ~ sample + patient * exon + condition * I(exon == exonID)

myExons <- testForDEU(myExons, formula0=formula0, formula1=formula1)

Best wishes,
Simone