[BioC] DESeq2 - interaction

[Michael Love]

hi Mike,

On Sun, Jul 20, 2014 at 10:11 AM, Mike Miller <mike.bioc32 at gmail.com> wrote:
> Dear All,
>
> I have problems in understanding what is the interpretation of default
> output of interaction contrast and how can I change the default settings
> (resultsNames(results)) and explore other interactions?
> For example, here are the 3 factors I am interested in:
> 1. treatment; levels: disease and control
> 2. localization; levels: A and B.
> 3. sex; levels: male and female
>
> If the design1 is: ~sex + treatment*localization, by default, the reported
> contrast in the results is:
> Wald test p-value: treatmentdisease.localizationA.
> Q1: Can I interpret the output as a list of genes that are diff.expressed
> (DE) between localization A and localization B in a disease?
> Q2: How could I get the DE genes between localization A and localization B
> in a control? Or DE genes between disease and control for localization A?
>

For the contrasts you are interested in performing, I would recommend
setting up the analysis as follows (I am planning to include such an
example in the vignette for the next release):

# create a new variable which combines treatment and localization:
dds$condition = factor(paste0(dds$treatment, ".", dds$localization))
design(dds) = ~ sex + condition
dds = DESeq(dds)
# then for example, DE genes between localization A
# and localization B in a control
res = results(dds, contrast=c("condition","control.B","control.A"))
# or DE genes between disease and control for localization A
res = results(dds, contrast=c("condition","disease.A","control.A"))

Mike

> If I change the order of factors in a design:
> design2=  ~sex + treatment*localization, then by default in the results I
> get
> Wald test p-value: localizationA.treatmentdisease. Can I interpret this as
> a list of DE genes between disease and control in localization A? I tried
> it, but the results were the same as ones got from design1.
>
> Thank you in advance for your help.
> Best,
> Mike
>
>         [[alternative HTML version deleted]]
>
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at r-project.org
> https://stat.ethz.ch/mailman/listinfo/bioconductor
> Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor