Question: Limma Analysis, BH Correction - adj. P-value all the same value
0
gravatar for Cale Reid
3.5 years ago by
Cale Reid0
San Francisco, CA
Cale Reid0 wrote:

Hi All,

I'm new to R and Limma, so apologies in advance for my naivete.

I'm analyzing some microarray data, and after applying the BH correction w/ topTable, I end up with a list of genes where the adj. P-value is apparently all the same. How do I interpret that? The P-values (unadjusted) vary significantly, but the ad. P-value for each gene is the same number (0.2800681 in my case).

I also note that some of my top results are labeled as normgene->intron (aka not real genes). Could this be the issue? I used the oligo package for initial processing of my data.

Any guidance would be much appreciated.

limma R • 2.7k views
ADD COMMENTlink written 3.5 years ago by Cale Reid0

If you make a histogram of the unadjusted p-values, what does the graph look like? Just upload that image somewhere and provide a link.

ADD REPLYlink written 3.5 years ago by Devon Ryan91k

Hey Devon. Thanks for your help. A histogram of the unadjusted P-values looks like this: http://imgur.com/SBJf1zu

ADD REPLYlink written 3.5 years ago by Cale Reid0

You'll need to show all of the genes, not just the top 100.

ADD REPLYlink written 3.5 years ago by Devon Ryan91k

Done: http://imgur.com/OpO1eiK

top_100 is a bit of a red herring - the 100 refers to my experimental condition, not the # of genes, but yeah - I definitely hadn't included all genes in the histogram before

ADD REPLYlink written 3.5 years ago by Cale Reid0

Wow, that looks like you should have some significant genes after correction. Do you get the same results if you manually run p.adjust() on the p-values?

ADD REPLYlink written 3.5 years ago by Devon Ryan91k

Here's what my table looks like after using p.adjust manually: http://imgur.com/GXXmpdw

P-value and adjusted p-value are from limma. The actual_adjusted p-value is from taking the limma provided p-value and using p.adjust with the BH correction.

Any thoughts? I can't figure out why they'd be different even though supposedly both limma and p.adjust are using BH.

ADD REPLYlink written 3.5 years ago by Cale Reid0

That's by far the craziest (bioinformatics related) thing I've seen this week. I'm pretty sure limma is even calling p.adjust() internally, so I have absolutely no clue why it's giving you those results. You might post this over on the bioconductor site, where Gordon Smyth (the limma author) can have a look.

ADD REPLYlink written 3.5 years ago by Devon Ryan91k

I posted on the bioconductor website - thanks again for your help. This is mystifying my partner and I as well so hopefully someone will know what's up!

ADD REPLYlink written 3.5 years ago by Cale Reid0

Would have been helpful to provide the link to your post on bioconductor's webpage.

ADD REPLYlink written 2.3 years ago by sina.nassiri40

Found it:
https://support.bioconductor.org/p/79397/

ADD REPLYlink written 19 months ago by emyers0
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 639 users visited in the last hour