[BioC] DEXSeq filtering

[Simon Anders]

Hi Mali

On 2012-05-31 07:47, mali salmon wrote:

> Two questions regarding DEXseq
> 1. Would you recommend to do an independent filtering prior to differential
> exon expression analysis (similar to what is done with gene expression
> analysis)?

Use the parameter 'minCount' in estimateDispersions for this. The 
default, 10, is probably reasonable, but maybe a bit on the low end. 
Every exon with less than 10 counts (summed across all sample) is 
omitted from the tests.

> 2. What about exons with 0 counts across all samples, does DEXSeq uses them
> when estimating overall gene expression?  would you recommend to remove
> those before analysis?

No need to remove them manually, estimateDispersion does this anyway.

   Simon