betaPrior and lfcShrink
1
0
Entering edit mode
@35aa14a7
Last seen 27 days ago
United States

Hi,

I am using DESe2 (v 1.33.4) to compare three shade avoidance phenotype in Arabidopsis.

I observed that the betaPrior in current version of DESeq2 was set to FALSE. If this is the case, we can compare C vs A and B vs A but we can't compare C vs B group.

When I tried setting the betaPrior to TRUE, I observed that I can get a group of A, B, C and I can make the three contrasts thereafter but I cannot use the lfcShrink function.

First question, how do you get the C vs B with betaPrior still set to FALSE and still be able to use the lfcShrink function?

Second question, if I want to detect WEAKLY EXPRESSED GENES between each group, would it be better to set the betaPrior into TRUE so I can ran the contrasts argument and alongside the lfcthreshold argument?

DESeq2 • 316 views
1
Entering edit mode
@mikelove
Last seen 1 day ago
United States

With default DESeq2, you can just do:

dds <- DESeq(dds)
resCB <- results(dds, contrast=c("condition","C","B"))


For more details see ?results, and the vignette:

https://bioconductor.org/packages/release/bioc/vignettes/DESeq2/inst/doc/DESeq2.html#contrasts

Then to shrink the CB effect, you can either use type="ashr" for shrinkage (easy), or for type="apeglm" you can do one extra step to make B the reference level and re-run just nbinomWaldTest() on the data to make C vs B one of the coefficients listed by resultsNames() (just two extra lines of code).

This is described in the "Extended section on shrinkage estimators" in the vignette.

What is a "weakly expressed gene between each group" I don't follow. Maybe give an example of what counts would look like.