Question

compare one group to mean of two others in DESeq2 with betaprior=FALSE

0

Entering edit mode

Chris Stubben ▴ 40

@chris-stubben-12524

Last seen 4.1 years ago

Salt Lake City, Utah

Is it possible to compare one group with the mean of two others if betaPrior=FALSE? In previous versions of DESeq2, I could group C vs A+B using a list below, but I'm not sure how to work with the new resultsNames output like trt_B_vs_A.

dds <- makeExampleDESeqDataSet(n=1000, m=18)
dds$trt <- factor(rep(c("A","B","C"),each=6))
design(dds) <- ~ trt
dds <- DESeq(dds, betaPrior = TRUE)
resultsNames(dds)
[1] "Intercept" "trtA"      "trtB"      "trtC"    

res <- results(dds, contrast = list("trtC", c("trtA","trtB")), listValues=c(1, -1/2))

dds <- DESeq(dds)
resultsNames(dds)
[1] "Intercept"  "trt_B_vs_A" "trt_C_vs_A"

deseq2 • 1.2k views

ADD COMMENT • link updated 6.2 years ago by Michael Love 41k • written 6.2 years ago by Chris Stubben ▴ 40

score 1 · Accepted Answer · 2018-02-06

1

Entering edit mode

Michael Love 41k

@mikelove

Last seen 15 hours ago

United States

You can get the same comparison with ~0 + trt.

ADD COMMENT • link 6.2 years ago Michael Love 41k

0

Entering edit mode

Thanks, I remember reading that you should not use ~0 + trt, but that must be when betaPrior=TRUE

ADD REPLY • link 6.2 years ago Chris Stubben ▴ 40

0

Entering edit mode

Yeah, DESeq2 will actually stop you from using ~0 + trt when betaPrior=TRUE

 Error in designAndArgChecker(object, betaPrior) :
   betaPrior=TRUE can only be used if the design has an intercept.
   if specifying + 0 in the design formula, use betaPrior=FALSE

ADD REPLY • link 6.2 years ago Michael Love 41k

0

Entering edit mode

Is there any way to shrink fold changes if you use ~0 + trt? With betaPrior = FALSE, I get the same error at a later step running lfcShrink(dds, contrast= c("trt", "C", "A"))

ADD REPLY • link 6.2 years ago Chris Stubben ▴ 40

0

Entering edit mode

The easiest way to get this would be to use betaPrior=TRUE. This will always be supported. Or you can build the results table and then pass res to lfcShrink with type=“ashr”.

ADD REPLY • link 6.2 years ago Michael Love 41k