Kernel Machine--Based Testing with Paired Genetic Samples

• Activity Number: 651
• Type: Contributed
• Date/Time: Thursday, August 13, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Yatong Li* and Michael C. Wu
• Companies: University of Washington and Fred Hutchinson Cancer Research Center
• Keywords: kernel machine ; SKAT ; composite kernel ; paired data

Abstract

Kernel machine methods, such as the SNP-set/sequence kernel association test, have become a popular strategy for assessing the association between multiple genetic markers in a gene region or a pathway and a complex trait. Despite its popularity, the SKAT approach is not appropriate for the analysis of paired genetic samples, e.g. mother-child pairs or donor-recipient pairs, in which there is a single outcome for each pair of genetic samples. Therefore, working with the kernel machine testing framework, we consider a new strategy for testing the association between the outcome and pairs of genetic samples at the pathways level. Specifically, as in the SKAT, we use a regression model in which genetic variants are modeled parametrically or non-parametrically using kernel machines. Our framework particularly models the main effects of the genetic markers in each individual using a composite kernel. We show by simulation studies that the test has correct size and reasonable power across a wide range of scenarios. We also illustrate the method via application to some real genetic association studies with paired data.