Is the same order when using p value or q value for compareCluster analysis?

Question

[Lu.Lu]

Hey,

I'm so confused about a simple question, hope you guys could help me!

So I do a GO analysis using compareCluster as I compared my down-regulated genes and up-regulated genes. I'm wondering whether the order is the same if I use p value or q value to plot the GO results? or the basic question is the rank order the same when use p value or q value?

This pictures I plotted using dotplot function, and ggplot2 separately.

Answer 1

[seidel]

Adjusting P-values inflates the values but does not change the order of the values. You can test this by plotting the the P-values and the Q-values against each other. In cases where some sets of P-values are identical, it's not clear how to control the sorting order. Are you explicitly determining the order in each of your plots?

[ATpoint]

Since qvalues/FDR can have ties (for example large pvalues can often have FDR of 1) it would depend on how ties are broken by the ordering method. I would order by pvalue but report FDR/q.

[Lu.Lu]

yes, you are right! I also found that, the large p values have FDR of 1.

[Lu.Lu]

Thanks so much for your kind reply! I understand this now, but from my data, it seems they are different, the first one is sorting by p value, the second one is sorting by q value, look at "signal release".

so I was so confused!

and yes, I want to have a final plot that should be like this,

do you have some ideas to do that?