[BioC] Query regarding DESeq

Wolfgang Huber whuber at embl.de
Wed Jul 30 11:49:07 CEST 2014


Dear Gourja

> Do you think the results which I got after it will be significant?

The better question is whether they are useful. If I correctly understood what you propose to do, you will observe a lot of confounding [1], i.e. apparent differential expression in an isoform that really has nothing to do with biological regulation of that isoform, but with changes in abundance of other transcripts with which it happens to overlap.

Packages like DEXSeq, BITSeq, FlipFlop, edgeR::spliceVariants and many others to whose authors I apologise for not lising here were designed to overcome just that problem, with different tradeoffs. It’s worth having a look at their associated papers.

[1] see any stats book, or indeed http://en.wikipedia.org/wiki/Confounding

Wolfgang




Il giorno Jul 30, 2014, alle ore 2:03 EDT, Gourja Bansal <gourja.bansal at igib.in> ha scritto:

> Hi Simon,
> 
> I have a query regarding count based method to find differential expression.
> 
> As in seqanswers you stated that HTSeq is not for isoform level count as
> most of the reads are overlapping, so a major portion of data will be
> reagrded as ambigous.
> 
> But if instead of rejecting overlapping reads if I assign them to all
> isoforms in which they are falling i.e. by using featurecounts and then
> find differential expression using DESeq. Do you think the results which I
> got after it will be significant?
> 
> Please reply.
> 
> Regards,
> Gourja
> 
> 	[[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



More information about the Bioconductor mailing list