A Multivariate Approach on Genome-Wide Association Studies (GWAS) by Modeling Multiple Traits Simultaneously to Identify Pleiotropic Genetic Effects

• Activity Number: 104
• Type: Invited
• Date/Time: Monday, August 1, 2011 : 8:30 AM to 10:20 AM
• Sponsor: WNAR
• Author(s): Yi-Hsiang Hsu*+ and Xing Chen and David Karasik and Kathryn Lunetta and Douglas Kiel
• Companies: Hebrew SeniorLife Institute for Aging Research/Harvard Medical School and Harvard School of Public Health and Hebrew SeniorLife Institute for Aging Research/Harvard Medical School and Boston University and Hebrew SeniorLife Institute for Aging Research/Harvard Medical School
• Keywords: GWAS ; PLEIOTROPY ; SNP ; MUTIVARIATE ; CORRELATED PHENOTYPES

Abstract

Pleiotropy occur when multiple traits were affected independently by same genetic variants. Due to correlation among traits and moderate genetic effects of GWAS, it is inefficient to detect pleiotropy by univariate analytical framework. We propose here a new approach to test pleiotropy on GWAS using a two-stage strategy: in the first stage, we performed a multi-phenotype GWAS by modeling traits simultaneously using our newly developed empirical-weighted linear-combined test statistics (eLC); and then, we tested the pleiotropy using a simplified structure-equation-modeling on selected SNPs from the first stage. eLC directly combines correlated test-statistics to an overall association test with a weighted sum of univariate statistics to maximize the information obtained from each univariate analysis. Using GWA16 simulated dataset, our eLC approach has outperformed the simple look-up on the overlaps among univariate GWAS and other multivariate methods (such as MANOVA, GEE and PCA). We applied our approach to data from the CHARGE and GEFOS consortia and identified pleiotropic genetic effects on reproductive phenotypes (age at menarche and age at menopause) and bone mineral density.