Control of Population Stratification by Principal Components--Based Genomic Propensity Scores in Genome-Wide Association Studies

• Activity Number: 537
• Type: Contributed
• Date/Time: Wednesday, August 7, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Huaqing Zhao*+ and Nandita Mitra and Timonthy R. Rebbeck
• Companies: Temple University School of Medicine and University of Pennsylvania and University of Pennsylvania
• Keywords: genome-wide association studies ; principal components analysis ; population stratification ; propensity score ; testicular germ cell tumors ; Tracy-Widom statistic

Abstract

Genome-wide association studies (GWAS) are an effective approach for identifying common genetic variants associated with disease risk. However, GWAS may be biased due to population stratification (PS). The most widely used method to address PS is principal components (PCs) analysis (PCA, Price et al., 2006). PCA uses genotype data to estimate "axes of variation", which can be used to adjust for association attributable to ancestry. The main challenge lies in choosing which PCs to include as covariates. The original proposal suggested using the 10 PCs with the highest eigenvalues. Such selection is arbitrary, and most importantly, patterns of local linkage disequilibrium will cause the presence of "nuisance axes". One recommendation is to use the Tracy-Widom statistic to select significant PCs (Patterson et al., 2006). However, this approach may result in a liberal selection of PCs. To overcome these limitations, we estimate genomic propensity scores based on PCs to control PS in GWAS. Simulations demonstrate that our approach can adequately control bias due to PS and preserve type I error and power. We illustrate our approach in a case-control GWAS of testicular germ cell tumors.