Remove Linked Markers Before Combining Populations, Or Afterward

Question

[pufferfish]

Various population genetics analyses have a requirement that the provided genotypes are only of unlinked markers. This can be easily accomplished with PLINK using the --indep-pairwise flag.

When working with different populations, should one remove the linked markers per population, and then merge subsequently (losing any markers not shared), or can one merge all the populations and then filter for linked markers? Intuitively, it seems that merging beforehand might mask the effects of linkage within the populations.

PS. I guess this isn't a PLINK (or SNP) specific question, but I'm using it as the example

Answer 1

[Zev.Kronenberg]

I have run ADMIXTURE/STRUCTURE and we merged before pruning. You want to treat all you data the same so you don't have a population specific systematic error.

I'm sure there are arguments in the other direction, but this is what I have done for pigeons and we were happy with the results.

[Zev.Kronenberg]

I am guessing you are doing a STRUCTURE analysis