Hello,

I have a dataset wherein a couple of genes have a p-value and padj at 0. I see that this is a rounded p-value number since the actual computed p-value is less then the positive double-precision floating point value.

I was curious to learn how folks were to plot and present these genes conveniently?

Very best,

Franc

There may be a few general comments or answers for this one. I will start the ball rolling, as I had to deal with this exact situation for EnhancedVolcano.

My original idea was to convert p-value=0 to the machine-lowest value. On my now aging HP laptop running Ubuntu 16.04, this equates to a very healthy low number:

.Machine$double.xmin
[1] 2.225074e-308

options(scipen = 999)

.Machine$double.xmin
[1] 0.00000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000002225074

I realised quite quickly that this ended up making the plot look weird in most cases. So, I eventually moved to converting these to 10^-1 of the lowest non-zero p-value:

p = 0.02

p * 10^-1
[1] 0.002

I issue a warning when this happens.

The code in question is here: https://github.com/kevinblighe/EnhancedVolcano/blob/master/R/EnhancedVolcano.R#L271-L291

Kevin