same padj for all the genes after DEseq analysis

Question

[Emily]

Hi everyone,

I have done a pair comparison with DEseq to find differentially expressed genes between two samples, for each sample i hvae four replicates, for the DEseq result, i got exactly same padj value for all the genes and it is not significant, is this normal ? I feel it is weird, even if there is no significantly expressed genes, I don't think the padj shoulf be the same for all the genes. Does anyone know what is going on here?

Thank you in advance.

[Manvendra Singh]

I am curious that what fold changes value you are getting for these genes?

Are they also close to one?

Answer 1

[dariober]

If there is never or little evidence against the null hypothesis the fdr is going to be always close to one, with many values the same. For example:

set.seed(1234)
p<- runif(n= 10000, min= 0, max= 1) ## Random pvalues
padj<- p.adjust(p, method= 'fdr')
unique(padj) ## Only 63 instead of 10000!
 [1] 0.9933484 0.9952445 0.9523590 0.9934287 0.9973141 0.9950672 0.9850355
 [8] 0.9935054 0.9983892 0.9955913 0.9975077 0.9985928 0.9956708 0.9992315
[15] 0.9981005 0.9943607 0.9894110 0.9979089 0.9945797 0.9977378 0.9960884
[22] 0.9971430 0.9919923 0.9920665 0.9963164 0.9943261 0.9952540 0.9994030
[29] 0.9939164 0.9950848 0.9948379 0.9991311 0.9950674 0.9990905 0.9956615
[36] 0.9856121 0.9953103 0.9936643 0.9966662 0.9968053 0.9948980 0.9947535
[43] 0.9981661 0.9981403 0.9943537 0.9992011 0.9943544 0.9574976 0.9984791
[50] 0.9987469 0.9934477 0.9995938 0.9865497 0.9954706 0.9989503 0.9936114
[57] 0.9990756 0.9993953 0.9943536 0.9915178 0.9913084 0.9806490 0.9984822

Answer 2

[Michael Love]

Hi,

Here is my response to this question on the Bioconductor mailing list: https://support.bioconductor.org/p/51916/#51952

Essentially, the same padj value repeated for multiple genes is a normal consequence of the Benjamini and Hochberg adjustment.

[Emily]

gene    padj
a       0.999909
s       0.999909
d       0.999909
f       0.999909
g       0.999909
h       0.999909
j       0.999909
n       0.999909
b       0.999909
m       0.999909
n       0.999909
c       0.999909
z       0.999909
x       0.999909

Thousands of genes got the same padj (0.999909). It is almost eaqual to 1. So this is normal? Thank you very much.