Testing for the Effect of a Genetic Pathway in Longitudinal/Clustered Data Using Kernel Machine Regression

• Activity Number: 452
• Type: Topic Contributed
• Date/Time: Wednesday, August 3, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Arnab Maity*+ and Stacey Alexeeff
• Companies: North Carolina State University and Harvard University
• Keywords: Longitudinal/clustered data ; Kernel Machine regression ; Score test ; Gene-gene interaction

Abstract

There is a growing scientific interest to test for genetic effects on disease by considering a set of genes that may be on the same biological pathway. Genes within a pathway may interact in functional ways to influence the progression of disease. Kernel machine regression has been introduced as a way to model a pathway effect, either parametrically or nonparametrically. We consider kernel machine regression for the testing of a genetic pathway effect in the longitudinal/clustered data setting, where a continuous disease outcome is measured repeatedly, possibly over time, for each subject. We develop a score-based test statistic for testing the effect of the genetic pathway accounting for the within subject correlation in the outcome variable. In addition, we present a simulation study to investigate the power of the test for different correlation structures, and we compare its performance with the test without accounting for correlation.