GWAS in mixed populations

Question

[USA_225478]

Hi Biostar community,

I have a theoretical question about GWAS that I'm hoping someone can help me with.

I have combined data from 7 populations to perform a GWAS, and found two SNPs significantly associated with my trait of interest.

I accounted for relatedness between individuals in the model, as I would expect different allelic frequencies between them.

Looking at the significant SNPs detected, I realise that both SNPs are monomorphic in three of the populations.

Does this invalidate the results?

Should these SNPs be filtered out prior to performing the analysis?

Thanks!

[Asaf]

Why? Do you suspect there is a one and only SNP causing the phenotype in all populations?

[USA_225478]

I'm not sure what you mean by this. I found two significant SNPs and they were both monomorphic in three of the populations. If the SNP does not appear in three of the populations, but the trait appears in all populations, I don't see how I can trust that this SNP can explains so much of the variance. I can't be sure though so I'd like input from people with more experience.

[Asaf]

Well, most of the phenotypes are not strictly caused by one SNP or even one gene. If the SNP is not very common it might be monomorphic in a population.

Answer 1

[Lemire]

Your results might possibly be spurious due to population stratification. You have allelic frequency differences between your population, and if the trait distribution also vary between your populations, then your results will likely be confounded. Unless you accounted for the different populations, which you didn't mention.

Best to analyze your 7 populations separately and then do some trans-ethnic meta-analysis. Googling these terms will direct you to recent and relevant literature.