How to interpret DESeq2's padj after setting lfcThreshold and the log2FC is greater than lfcThreshold?
1
0
Entering edit mode
Pratik Mehta ▴ 10
@0512b16f
Last seen 8 weeks ago
United States

Hello Biocconductor community,

I have been wondering this for some time (and I think I can finally articulate it, somewhat, well enough to have an idea of what to ask)

How would I interpret DESeq2's padj after setting lfcThreshold and the log2FC is greater than lfcThreshold?


An example:

So if the results() I use within DESeq2 looks like this:

res <- results(dds, 
              contrast=c("Group", second_compar, first_compar),
              independentFiltering=TRUE, 
              alpha=0.05, 
              pAdjustMethod="BH", 
              lfcThreshold=1, 
              altHypothesis="greaterAbs")

This would mean:

Fold Change = (second_compar/first_compar)

so:

log2(FC) = log2(second_compar/first_compar)

Then, as above, if I set the lfcThreshold=1 (so an FC equal to 2 in log2(FC)=... so lfcThreshold=1 would mean the cut-off I am setting is log2(2)=1 ) and then an alpha=0.05. This means that the padj value for a given differentially expressed & significant gene as defined by a log2(2) =1 and an alpha=0.05.

For everything above, I am somewhat okay in my understanding...


Ambiguous territory:

This is where it gets into "ambiguous territory" for me. Could someone help clear my understanding of what happens when with the "cut-offs" set above, if Gene X that has for example a log2(10) = 3.321928 but I am using the cut-off of log2(2)=1 and alpha=0.05?

And then to bring it to the actual example above, for hopefully the sake of simplicity, if first_compar in the Fold Change above had a value of 1 and second_compar had a value of 10, so that log2(FC) = log2(second_compar/first_compar) = log2(10/1) = log2(10) = 3.321928 and then let me just give a padj=0.001 for this Gene X comparison.

If you're still with me after reading all that. Thank you : ) Pratik

DESeq2 • 167 views
ADD COMMENT
2
Entering edit mode
@mikelove
Last seen 13 hours ago
United States

It's not a "cutoff" per se, but it is used in computing the p-value, this is described in the 2014 paper.

ADD COMMENT

Login before adding your answer.

Traffic: 630 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