Double Least Squares Kernel Machine Score Test for Genetic Pathway Effect

• Activity Number: 611
• Type: Contributed
• Date/Time: Wednesday, August 7, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Xiang Zhan*+ and Debashis Ghosh
• Companies: Pennsylvania State University and Penn State University
• Keywords: Adaptive test ; Equivalent kernel ; Garrot kernel machine ; Kernel smoothing ; Reproducing Kernel Hilbert Space

Abstract

Pathway and gene set-based approaches are very popular in genome-wide association studies (GWAS) and gene-expression profiling studies for assessing how molecules are related to disease outcome. Since most genes are not differentially expressed, existing pathway tests considering all genes within a genetic pathway suffer from considerable power loss. Moreover, for a differently expressed pathway, it is of interest to select important genes that drive the effect of the pathway. In this article, we propose a double least squares kernel machine (DLSKM) score testing procedure, which can both select important genes within the pathway as well as test the overall genetic pathway effect. This adaptive score testing procedure is based on a least squares kernel machine (LSKM) framework (Liu et al., 2007). The DLSKM procedure provides power gains relative to the ordinary kernel machine score test as well as other methods in various simulation settings. In addition, we investigate some theoretical properties of LSKM-based estimators and evaluate the performance of our DLSKM score test using simulation studies and a real data example.