ERROR: P<1e-300

Question

ERROR: P<1e-300

jielab opened this issue 7 months ago · 3 comments

Hi, there:

If some SNPs in my input GWAS has P-value < 1e-300, LDSC would fail, complaining that P-value is out of range (0, 1]。But P-value < 1e-400 is NOT exactly 0.

The latest GWAS on Height with millions of samples actually yielded a lot of SNPs with P < 1e-300.

It would be great if LDSC could address this, instead of throwing out an error.

Thanks!

Jie

Answer 1 · 2024-05-05T20:28:42.000Z

@jielab One potential problem is that if you convert such a p-value to a z-score, you will get an infinite z-score.

If you have the original z-scores from the height GWAS, this should not be an issue.

Answer 2 · 2024-05-06T01:46:27.000Z

Thanks!

Do you mean that I should calcualte a Z-score, that is, BETA / SE, and include that Z-score in my input file to run LDSC?

Answer 3 · 2024-05-06T23:13:33.000Z

@jielab Correct. The expected input contains columns SNP, Z, and N.