[BioC] DESeq2 get coefficients for each variable in the design formula

[Michael Love]

hi Quynh,

The default pipeline is for testing individual coefficients, and the
code you have is fine, defining two sets based on the two tests, and
look at the union of these.

It is also possible to perform a statistical test of both at once,
using a likelihood ratio test. There is a section in the vignette
explaining the LRT.

This would look like (using dds to shorten your object name):

dds <- DESeq(dds, test="LRT", reduced=~1)
res <- results(dds)

You should then add the two log fold change columns from the results
tables you have above as columns to this new results table. (I'm still
working in the devel branch on making this part simpler, specifying
contrast LFCs to add to LRT results tables).

Mike

On Tue, Jul 1, 2014 at 2:03 PM, Tran, Nhu Quynh T <qtran1 at uthsc.edu> wrote:
> Hi,
>
> I'm working on a RNA-seq data set and would like to control for age or bmi in the model together with the disease status, which I am able to do.  My question is how can I get the coefficients for both age and the disease together?  what I did was
>
> #Adjust for age:
> cds.acromegaly.age = DESeqDataSetFromMatrix(countData=acromegaly.protein.coding, colData=acromegaly.mapping, design=~age+group)
> cds.acromegaly.age$age <- relevel(cds.acromegaly.age$age, "(0,40]")
> acromegaly.cds.age <- DESeq(cds.acromegaly.age)
> acromegaly.results.age <- results(acromegaly.cds.age)
>
> #Get the results for group (or disease):
> acromegaly.results.age <- results(acromegaly.cds.age)
>
> #Get the results for age:
> acro.age.effect <- results(acromegaly.cds.age, contrast=c("age","(40.60]", "(0.40]"))
> sum(acro.age.effect$padj<0.05, na.rm=TRUE)
>
> Then I merge those two results to get the genes that are affected by age or affected by group, and affected by both.   Is there another way? Is there coefficients for each of the independent variables.  I saw only log2FCs and p-values.
> acro.age.combined <- merge(acromegaly.results.age, acro.age.effect, by.x="row.names", by.y="row.names")
>
> Thanks,
> Quynh
>
> _______________________________________________
> 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