Very low p-values

Question

[am29]

I did a GWAS and got very low p-values for many of the SNPs (I used Bonferroni correction). I wanted to make a Manhattan plot in RStudio. However, it seems that R has a maximum value of around e-324 for plotting the p-values, and is not able to plot my values which are lower than e-500.

Is there any other program that could plot these very low p-values?

Answer 1

[Matthias Zepper]

Yes, all R platforms are conforming to the IEC 60559 (also known as IEEE 754) standard. The precision for doubles (the largest of the 3 types of numeric values) is 53 bits, which allows representing a range of absolute values from about e-308 to e+308. So your only option is to import the already log scaled values and directly plot those as if they were simple numbers.

I have to say that I find those values odd and never encountered similarly low ones, so I'd recommend double-checking. You could also just use cut() on them and create significance categories like (n.s., , , ), because there is not really a biological meaning in differences that low for p-values of single SNPs.

[am29]

Thanks for the answer, but I would need more help on this. For example, how to import already scaled values? Also, is there any program that could plot these low p-values, instead of R?