Can PCs from PCA of "Chip Genotypes" be used for post-imputation GWAS? [to adjust for population stratification]

Question

[Oak]

Hi all,

I have both "chip" and "post-imputation" genotype data. I want to use principal components (PCs) of population genotypes to adjust for population stratification during GWA.

Is there a case against using PCs obtained from PCA on chip genotypes when running post-imputation GWAS? Or would there be a benefit in doing PCA again on post-imputation genotypes?

Many thanks for sharing your thoughts beforehand,

Answer 1

[4galaxy77]

Chips (usually) only contain common variants (MAF > 5%). These common variants have a limited ability to capture fine-scale structure. Imputed datasets contain much rarer variants, which can capture recent fine-scale structure. So my intuition is that doing a chip PCA might not adequately capture all the structure from a GWAS on imputed markers.

Ref: https://elifesciences.org/articles/61548

[Oak]

This is an interesting point and thank you for sharing the publication. My impression is that MAF filtration is performed as part of QC steps on imputation data before PCA as well. Would not it affect the outcomes in a similar fashion?