Pruning a dense SNP dataset without losing too much LD

Question

[yorgos.athanasiadis]

Hi,

I have a set of ~1.7 million SNPs and I want to bring it down to ~300,000, but without losing too much LD between the remaining SNPs (because the methods I'm using after pruning rely heavily on IBD tracts).

What are the best parameters for the --indep-pairwise flag in PLINK to create a dataset with the desirable features? Any ideas?

Thanks!

Answer 1

[chrchang523]

If you’re explicitly trying to keep LD, and you’ve already performed basic quality control, —maf and —bp-space are probably the best tools to use.

Answer 2

[Sean Davis]

You may need to experiment with increasing the --indep-pairwise flag a bit to get down to 300k. There is really only one parameter to play with, so just keep going until you reach your desired density.