This is a test version of Biostars. For the public version, visit https://www.biostars.org.
FDR of fisher test

Hi all,

I have some human cancer RNA-seq data (control/treatment) , I did differential expression analysis (with CLC Genomics) and performed meta-analysis (fisher test) by metaDE package of R.

now I want to define a threshold for FDR, but after defining a threshold (FDR ≤ .05) Nothing left, because the smallest FDR is 0.1651751 and the largest is 0.9999993,and Most of the numbers is around 0.98.why are they so big?

rna-seq r
• 352 views • link
updated by dariober dariober • written by Edalat • Edalat

Removing background would give you better FDRs. Get rid of genes not expressed. Also, double check if you are talking about FDRs or P-values ;)

• 0 views • link
written by biotech biotech

1 answer

why are they so big?

Mathematically because FDRs are calculated from:

cummin(length(x)/(length(x):1) * x)

where x is the vector of p-values sorted largest to smallest and cummin is the cumulative minimum.

In practice, this means that none of the tested genes can be confidently claimed to be differentially expressed, even those with smallest p-values are compatible with the null hypothesis of no difference between control and treatment.

• 0 views • link
written by dariober dariober

thank you for reply.I think it is better to change method of meta-analysis.

• 0 views • link
written by Edalat • Edalat

Log in to answer this question.

More posts like this