[Bioc-sig-seq] problem when using DEXSeq on genes with 1 exon only

[Simon Anders]

Hi Jane

On 08/08/2011 02:45 PM, Wolfgang Huber wrote:
> in the case of genes (or better term: loci) with only one exon, DEXSeq's
> testGeneForDEU does not make sense - it is looking for exons within the
> locus that behave differently from the average of all the others: there
> are no others! Hence the parameters are unidentifiable, and the software
> should not even look at them.

Actually, it doesn't. This is why the p value is NA. Also note the 
column 'testable' in the fData data frame: it is set to FALSE in such 
cases to indicate that a test was not attempted. You will also find 
testable=FALSE occasionally in genes with more than one exons, namely 
whenever all exons or all exons but one have zero counts in all samples.

Cheers
   Simon