Adding lfcShrink to DESeq2 analysis

Question

[Trivas]

I'm trying to incorporate lfcShrink into my DESeq2 analysis and I'm running into a problem with the coef argument in lfcShrink.

I performed an experiment with 4 cell lines, treated and untreated, in biological triplicate, for a total of 24 samples. I can use

results(dds, contrast = c("condition", "cell.line.2.treated", "cell.line.2.untreated"))

and get my DGE results and downstream analysis is fine.

My problem arises when trying to use lfcShrink. This forces me to use the coef argument, but when I look at the options in resultsNames, all of the comparisons are with cell.line.1.untreated, e.g.

condition_cell.line.2.treated_vs_cell.line.1.untreated
condition_cell.line.2.untreated_vs_cell.line.1.untreated
condition_cell.line.3.treated_vs_cell.line.1.untreated
condition_cell.line.3.untreated_vs_cell.line.1.untreated

which isn't what I need for my analysis.

I've tried looking into setting the base condition in the DESeq object without any luck. The only thing I can think of doing is setting cell.line.2.untreated as the first level within my colData object, but I'm thinking there's a better way of doing this.

I tried running lfcShrink with contrast instead of coef and get this error:

Error in lfcShrink(dds, contrast = c("condition", "cell.line.2.treated", "cell.line.2.untreated"), : 
type='apeglm' shrinkage only for use with 'coef'

[rpolicastro]

To my knowledge setting the factor level is the only way to accomplish this with apeglm.

[jv]

I've tried looking into setting the base condition in the DESeq object without any luck.

use relevel to specify the reference level for this factor: http://bioconductor.org/packages/devel/bioc/vignettes/DESeq2/inst/doc/DESeq2.html#note-on-factor-levels

or use ashr shrinkage method which I think uses contrast instead of coef.

For more info see http://bioconductor.org/packages/devel/bioc/vignettes/DESeq2/inst/doc/DESeq2.html#extended-section-on-shrinkage-estimators

Answer 1

[ATpoint]

See here: type='apeglm' shrinkage only for use with 'coef'