Comparisons of SNP Partitioning Strategies for Regression-Based Multi-Marker Scanning of Genetic Association

• Activity Number: 623
• Type: Contributed
• Date/Time: Thursday, August 2, 2012 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Yun Joo Yoo*+ and Shelley B. Bull and Lei Sun
• Companies: Seoul National University and Samuel Lunenfeld Research Institute and University of Toronto
• Keywords: Multi-marker test ; Genome-wide association study ; Multiple hypothesis testing ; Gene-based test ; SNP ; Power comparison

Abstract

Multi-marker tests for genetic association data of single nucleotide polymorphisms can be more powerful than individual SNP based tests when causal effect are captured by multiple SNPs and aggregated into a global test. When applied to genomic scanning, however, researchers need to decide on the method of grouping or partitioning SNPs In this study, we compare several different strategies regarding group size and the choice of partitioning or sliding windows for several regression-based multi-marker tests in terms of power and type I error. We used a real data set from HapMap to simulate a region of genetic data with realistic linkage disequilibrium pattern and obtained covariance matrices of multiple multi-markers tests under null hypothesis. Using the estimated covariance matrix, the power of different strategies are obtained and compared after adjusting type I error inflation. Among various grouping strategies, a binning method using correlation between SNPs performed well combined with multi or single degrees of freedom linear combination tests and proper allele coding methods.