Proportion testing for linkage in genomic studies of cosmopolitan populations

Authors

• K Morisaki
• Q Yuan
• C Hodgkinson
• D Goldman

Abstract

Genomic approaches are highly successful in linking genes to phenotypes. However, a longstanding problem is ancestral variation in allele frequencies, leading to both inflation and deflation of p-values whenever cases and controls mismatch in ancestry. Principal Components are commonly used as covariates to correct for ancestry, only partly overcoming mismatch as shown by analyses of phenotypes, such as stature, that show strong ancestral variation. The effects of stratification may be worse for rarer genotypes. Here, we report a new method to correct for ancestry by comparing cases and controls in study samples against large numbers of population controls such as gnomAD. We analyzed genotype frequencies of study cases and controls against expected proportions. Using genotypes from a large, multi-ethnic sample of AUD patients with and without alcohol hepatitis (AH), frequencies of most genotypes, not linked to AH, were well-predicted by the binomial distribution, allowing for usage of separate two-sample proportion tests. However, highly significant (p<10E-10) genotype-based linkage of the known AH gene PNPLA3 was observed. The inter-correlated statistics for the three genotypes observed at biallelic loci may later be combined, and the significance computed based on the distribution of a linear combination of chi-squared distributions.

Scientific Focus Area: Genetics and Genomics

This page was last updated on Tuesday, August 6, 2024