limma decideTests: only p-value less than (but not equal to) alpha are flagged as significant
1
1
Entering edit mode
@robertospreafico-7544
Last seen 7.1 years ago
United States

Hello,

I realized that the decideTests function in limma selects as differentially expressed only genes whose p-value is less than an alpha threshold, rather than those that are less than or equal to alpha. Taking this to an extreme case, when alpha = 1, some genes are still not selected as differentially expressed. Is there a reason for this? Would it make sense to add a parameter to the function to define whether the alpha threshold is inclusive or exclusive (with default maybe being inclusive)?

Thank you for your suggestions!

Best,

Roberto

limma edger significance rnaseq • 973 views
ADD COMMENT
0
Entering edit mode

decideTests uses a <= operator to select DE genes with adjusted p-values below the supplied p.value, so I'm not sure how you're getting the behaviour you observe. Please provide a minimum working example to demonstrate.

ADD REPLY
2
Entering edit mode
@gordon-smyth
Last seen 19 minutes ago
WEHI, Melbourne, Australia

If you reinstall limma, you will find that decideTests() was modified a few months ago so that it does in fact select genes with p-values less than or equal to the specified level. Here is a simple example that you can run for yourself:

> example(lmFit)
> summary(decideTests(fit2,p=1))
   Grp1 Grp2vs1
-1   52      52
0     0       0
1    48      48

As you can see, all genes are DE.

However I have never seen a data example for which "<" vs "<=" would make any difference except for p=1. It is very unlikely indeed that a p-value would be exactly equal to 0.05, or any other specified value, in floating point arithmetic. So I don't see a compelling reason to offer options in this regard.

 

ADD COMMENT
0
Entering edit mode

Thank you! We use R 3.2 in production, and the latest limma available for that is 3.26.9 - Will try to have the infrastructure updated to 3.3/Bioconductor 3.4 to take advantage of the latest limma.

Thanks!

ADD REPLY

Login before adding your answer.

Traffic: 704 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6