[BioC] multifactorial analysis

[Simon Anders]

Hi Andreia

On 2012-04-09 13:06, Andreia Fonseca wrote:

> I am planning an experiment to profile miRNAs by ngs with the following
> multifactorial design: breed, tissue type, diet, tissue color. Is it
> possible with DEXSeq to account with the effect of interaction between
> these factors?

Sure, however you want to use DESeq, not DEXSeq. The latter is to test 
for differential exon usage, and as there is no alternative splicing of 
miRNAs, this is not what you want. DESeq allows you to fit a GLM on 
expression levels and test for main or interaction effects.

By the way, I hope you realize that a design with so many factors is 
challenging and requires a large number of samples, especially given 
that you will have to cope with differences in genetic background within 
breed, which will be hard to control for. Assuming you are talking about 
mammals, I would be rather surprised if it turned out possible to get 
enough power to detect interactions without a sample number going into 
the hundreds.

   Simon