[BioC] Deseq for pairwise comparison among many pairs

Michael Love michaelisaiahlove at gmail.com
Wed Jun 11 15:32:11 CEST 2014

hi Shihai,

We generally recommend users upgrade to DESeq2, which is faster and
more sensitive. Additionally, it would make the pairwise comparisons a
bit easier. You could use for example:

dds <- DESeq(dds)
results(dds, contrast=c("treatment","B","A"))
results(dds, contrast=c("treatment","C","A"))
results(dds, contrast=c("treatment","C","B"))

Which would estimate size factors and dispersions using all samples in
the first line. The following three lines build results tables
contrasting all pairs of the three treatments.

I would guess you can run them together, as genes from different
chromosomes are combined in a standard analysis. But if I weren't
sure, you could compute the size factors from the chromosomes and from
the plasmids separately. If these are similar, then I would run them
all together.

If you have two vectors which identify the plasmid rows and the chromosome rows:

estimateSizeFactorsFromMatrix( counts(x)[ plasmidRows, ] )
estimateSizeFactorsFromMatrix( counts(x)[ chromosomeRows, ] )

where x is the dataset, either CountDataSet or DESeqDataSet for DESeq2.


On Tue, Jun 10, 2014 at 3:14 PM, Feng, Shihai <sfeng at lanl.gov> wrote:
> I have two  questions related to how to use Deseq.
> 1) I have total three sample treatments. I am doing pairwise comparison between treatments. I am wondering what sample pool I should use to calcualte SizeFactors and Dispersion.   Should I  use only  the two treatments data that I use in the subsequent nbinomTest  step or use all my three sample treatments to calculate SizeFactors and Dispersion.
> 2) My sample is bacterial that includes chromosome and plasmids, should I study them separately or together ? More specifically,  should I run your deseq pipeline independently on chromosome and plasmid or should I combine  chromosome and plasmid data first then run your deseq pipeline.
> thanks,
> shihai
> _______________________________________________
> 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