Base-10 log transformed P-values and clump with plink v.1.9

Question

[docj]

Hi.

Assume I have a list of SNPs and their respective P-value on the -log10 scale. To find the lead variant(s) using --clump, I need to supply plink (v.1.9) with a data file which contains the IDs of the SNPs and their P-value. As P = 10^{-log10(P)} we will run into precision problems if -log10P > 323 (at least in R). After some searching it seems that it isn't possible to tell plink to use the log-transformed P-values.

From the documentation, it seems that --clump selects the smallest P-value at each iteration. Could I simply truncate any log-transformed P-value which exceeds the precision threshold such that the order is preserved, even if the P-values themselves are "bogus"? Or is there a better way to get around this problem?

Thanks.

Answer 1

[chrchang523]

Yes, that is the standard plink 1.9 --clump workaround for tiny p-values.