Question

Error in lfcShrink

0

Entering edit mode

Claudio Carril • 0

@claudio_carril-24100

Last seen 3.9 years ago

University of Liège, Belgium

Dear community I am writing to you because I have a problem with DESEq2 when I try to use lfcShrink. After that, it is not possible to continue with the analysis

This error started to happen when I run my analysis in Ubuntu 20.04.1 and use R 4.0.3, when I ran the same analysis in the previous version of R there was not a problem

The line:

> resShrunk <- lfcShrink(dds, contrast = c("condition", "3N_D", "20C_G2"), cooksCutoff=F)

An the error:

> resShrunk <- lfcShrink(dds, contrast = c("condition", "3N_D", "20C_G2"), cooksCutoff=F)
Error in lfcShrink(dds, contrast = c("condition", "3N_D", "20C_G2"), cooksCutoff = F) : 
  type='apeglm' shrinkage only for use with 'coef'

Thanks in advance

lfcShrink DESeq2 • 3.3k views

ADD COMMENT • link updated 4.0 years ago by tcalvo ▴ 100 • written 4.0 years ago by Claudio Carril • 0

1

Entering edit mode

You can either use coef or you can switch back to type="normal' which was the previous default value for type.

The reason why it depends on the version, is that type="normal" was deprecated over two release cycles, and now is no longer default (this was printed as a message to the console for two releases, before the change was made). It performs worse than apeglm/ashr. Changes are documented here. See description of shrinkage estimators in the vignette for more details.

ADD REPLY • link 4.0 years ago Michael Love 42k

0

Entering edit mode

Thank you so much for your help!

Now it is working properly

Best regards

ADD REPLY • link 4.0 years ago Claudio Carril • 0

score 1 · Answer 1 · 2020-11-09

1

Entering edit mode

tcalvo ▴ 100

@tcalvo-12466

Last seen 16 months ago

Brazil

Hello, Claudio.

lfcShrink does not support apeglm + contrast. You should use the coef argument instead of contrast when using apeglm. This will solve your issue.

ADD COMMENT • link 4.0 years ago tcalvo ▴ 100

0

Entering edit mode

Thank you, I see that

But that with that I have more question about how to modify the coef to use the apeglm. Because, the coef it does not give me the combinations I am looking for

but for that I am going to open another question in the forum

best regards

ADD REPLY • link 4.0 years ago Claudio Carril • 0

0

Entering edit mode

Sorry, been absent from here. Did you find a solution? I usually just change the levels of my factor variable, usually the base level to solve the issue. Probably there are better ways to do it.

ADD REPLY • link 4.0 years ago tcalvo ▴ 100