Statistical Issues Arising from Analysis of Genome-Wide Case-Control Association Studies

• Activity Number: 39
• Type: Invited
• Date/Time: Sunday, July 29, 2007 : 4:00 PM to 5:50 PM
• Sponsor: ENAR
• Author(s): Gang Zheng*+ and Jungnam Joo and Colin O. Wu and Nancy L. Geller
• Companies: National Heart, Lung, and Blood Institute and National Heart, Lung, and Blood Institute and National Heart, Lung, and Blood Institute and National Heart, Lung, and Blood Institute
• Keywords: Quality control ; Multiple testing ; Population structure ; Replication ; Robust test

Abstract

We discuss several statistical issues arising from our analysis of a genome-wide association study (GWAS) using over 100K SNPs. In quality control, we discuss the rational of testing HWE and a possible trade-off between high per SNP call rate and high per SNP error rate. In genome-wide scans (GWS), at least three common, simple test statistics can be used. We propose a robust tag statistic. Multiple testing is a serious issue in analysis of GWAS, starting even with quality control. We discuss when one should correct the multiple testing. In using genomic control or delta centralization to correct for cryptic relatedness and population stratification, null loci are used. Issues of how to choose null loci will be discussed. Finally, we discuss issues in replicating results from a GWS, including sample size calculations and how to set the significance level for replications.