The Generalized Higher Criticism for Testing SNP-Sets in Genetic Association Testing

• Activity Number: 514
• Type: Contributed
• Date/Time: Wednesday, August 6, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Ian Barnett*+ and Rajarshi Mukherjee and Xihong Lin
• Companies: Harvard and Harvard and Harvard School of Public Health
• Keywords: Higher criticism ; Signal detection ; Multiple hypothesis testing ; Correlated test statistics ; Genetic assocition testing

Abstract

Genes, gene pathways, and network effects can contribute to the risk of complex genetic diseases. These genetic constructs each contain multiple SNPs, and only a sparse subset of these SNP-sets are generally related to the disease of interest. In this paper we adapt the higher criticism, a test traditionally used in high dimensional signal detection settings, to genetic association testing for SNP-sets. The higher criticism performs well when the signal is sparse, making it a potentially powerful tool for SNP-set association testing. Unlike past treatments of the higher criticism, we propose the generalized higher criticism (GHC) that does not require asymptotics in the number of SNPs in the SNP-set while simultaneously allowing for arbitrary correlation structures among the SNPs in the SNP-set. The power of this method is compared with existing SNP-set tests over simulated regions with varied correlation structures and signal sparsity. The relative performance of these methods is also compared in their analysis of the CGEM breast cancer genome-wide association study.