Functional Genome-Wide Association Studies Using Sparse Group Lasso

• Activity Number: 127
• Type: Contributed
• Date/Time: Monday, August 5, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Qing Pan*+ and Yunpeng Zhao
• Companies: George Washington University and George Mason University
• Keywords: Diabetes complications ; Functional data analysis ; B-spline smoothing ; Group variable selection ; Genome wide association study ; Sparse group lasso

Abstract

In GWAS, the one SNP at a time testing strategy fails to consider the effect of multiple SNPs simultaneously and is subject to severe multiple comparison adjustments. Furthermore in complex diseases, genetic effects often changes over time. In cases of phenotypic data collected at various time points, single time point methods do not utilize information available in data effectively. Hence models incorporating the dynamic pattern of genetic influences over a time course are desirable. We develop a novel nonparametric model that incorporates multiple SNPs simultaneously with a continuous trait of interest measured at irregularly spaced time points. Specifically, we consider B-spline polynomials to approximate time-varying effects and use group lasso, which introduces both group-wise and within-group sparsity. The selection of important SNPs corresponds to the selection of groups of spline coefficients. The statistical properties of the proposed method are investigated through simulation studies. We apply out method to the GWAS of the Epidemiology and Intervention of Diabetes Complication trial where Type 1 Diabetes patients are followed for up to 27 years and annual GFR are taken.