Question: The FDR and P value in RNA-seq analysis
gravatar for mikysyc2016
2.5 years ago by
mikysyc201680 wrote:

Hi all, After using edgeR to do RNA-seq analysis, I get p valve and FDR. Which one you will use to select your interest target? I search website more people will recommend FDR value as the cutoff standard. But when read papers, a lot of paper use p value( like -log10P) as y axis to make volcano plot. So what is your suggestion? Thanks!

rna-seq p value and fdr • 4.9k views
ADD COMMENTlink modified 2.5 years ago by Dattatray Mongad350 • written 2.5 years ago by mikysyc201680

Volcano plot is the representation of effect (fold change) versus significance (-lgo10p-value). A very simple volcano could be like this but if you want to include FDR you can do it just like this.

ADD REPLYlink written 2.5 years ago by Nitin Narwade450

@OP: Please read more about p and FDR value calculation. FDR values are are adjusted p-values. People rely on adjusted p-values than p-values. However, there is another component in analysis that you can use for filtering: fold change (log fold change). However, both filters are arbitrary and please choose the cut off with the help of some one who knows /did the analysis earlier and validated the results.

ADD REPLYlink written 2.5 years ago by cpad011214k
gravatar for Dattatray Mongad
2.5 years ago by
National Centre for Cell Science, Pune
Dattatray Mongad350 wrote:

Mostly LogFC and -log10(pvalue) are used to draw volcano plots which are nothing but scatterplots. By these plots you can make illustrations that - how much of your transcripts/genes are significantly upregulated or downregulated. Here, p-values serve the purpose of checking significance.

p-value is a measure of how likely you are to get this data if no real difference existed. So, small p-value means your results are not by chance. But q-values are adjusted p-values by FDR approach, which show you even if your data is statistically significant, what would be False Discovery Rate. So, lower q-value is significant. Usually, if you sort p-value that order is maintained in q-value also.

So, if you want you can plot all transcripts (Log2FC on x-axis and -log10(p-value) on y-axis).

  1. p-value<0.05 => all significant
  2. log2FC>abs(1.5) => all differentially expressed
  3. padj<0.05 and log2FC>abs(1.5) => significant and differentially expressed

his would help reviewer. But use q-value as when you want to build some hypothesis on basis of them.

ADD COMMENTlink written 2.5 years ago by Dattatray Mongad350
Please log in to add an answer.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 1626 users visited in the last hour