How to find significantly stably expressed miRNA with DESeq2 DE results
3
1
Entering edit mode
mariakarand ▴ 20
@mariakarand-8971
Last seen 6.2 years ago
Norway

Hi,

I'm working with DESeq2 DE results of microRNA expression (which I have not myself performed). I want to identify both differentially expressed miRNA and significantly stably expresssed miRNA. The DE miRNAs are obviously the miRNA with p<0.05. How can I identify stably expressed miRNAs from the DESeq2 DE resaults? Is there any specific p-value threshold I can use (e.g all miRNA with baseMean over 10 and a p-value over 0.5 can be viewed as significantly stably expressed)?

Just to clearify, I'm not looking for stably expressed miRNAs to normalize the data, but rather to identify which miRNAs are DE, which miRNA are significantly stably expressed and which are neither.

-Maria

1
Entering edit mode
@sean-davis-490
Last seen 6 weeks ago
United States

I wrote a blog post summarizing some email exchanges on this subject.

http://watson.nci.nih.gov/~sdavis/blog/testing-for-non-differentially-expressed-genes/

0
Entering edit mode

Thanks!

A follow-up question: Is there any key to find what threshold you should use, or do you generally recomend a treshold -value of abs(0.3)?

And in the R-script it's first written

theta <- .3

But the "theta" is never used again, rather "thr" is used in the Z-tests. Is this a typo? Is "thr" really "theta"?

Maria

1
Entering edit mode
@mikelove
Last seen 1 day ago
United States

DESeq2 has built in functionality for testing for effects such that |beta| < T for some threshold T.

see the lfcThreshold and altHypothesis arguments in ?results.

There is a diagram in the vignette showing this, and a Figure in the DESeq2  paper showing an MA plot for a test of |beta| < T

http://www.genomebiology.com/2014/15/12/550/figure/F4

0
Entering edit mode
mariakarand ▴ 20
@mariakarand-8971
Last seen 6.2 years ago
Norway

Thanks!

A follow-up question: Is there any key to find what threshold you should use, or do you generally recomend a treshold -value of abs(0.3)?

And in the R-script it's first written

theta <- .3

But the "theta" is never used again, rather "thr" is used in the Z-tests. Is this a typo? Is "thr" really "theta"?

Maria