Figure 1.

Effect Explained by Local Ancestry and an eQTL Association Test

We evaluated the effect explained by local ancestry and the correlation between local ancestry and genotype.

(A and B) An empirical distribution of the maximum absolute-effect size of local ancestry for each gene-expression trait (betaLA) in the NIGMS dataset (admixed samples, A) and in the GTEx whole-blood dataset (multiethnic samples, B), showing a large effect for a substantial number of genes.

(C) A comparison of the genotype-local-ancestry (LA) correlation and genotype-principal-component (PC) correlation in the NIGMS dataset. The distribution for LA is skewed to the right (or higher values of the correlation), indicating that multi-collinearity, and thus inflated variance of estimated SNP effect size on gene expression [as quantified by the variance inflation factor of the ancestry predictor, $VIF\left(ancestrypredictor\right)=1/\left(1-R^2\right)$], is a greater problem for LA than for PC.