Abstract MP18: Examination of Genetic Associations with Individual Lipoprotein Subfraction Measures, Aggregated Across Multiple SNPs, in Four Ethnic Groups within the Multi-Ethnic Study of Atherosclerosis (MESA)

Abstract

Introduction: A large GWAS conducted in individuals of European descent identified several genetic markers (SNPs), subsequently validated as being associated with fasting lipoprotein measures (Chasman et al, 2009, PloS Genetics ). We examined the reported associations separately in the Caucasian- African-, Hispanic- and Chinese- American MESA participants. Hypothesis: Genetic variants associated with lipoproteins in Caucasians will not be associated across all ethnic groups. Methods: Lipoproteins were determined by nuclear magnetic resonance spectroscopy in 4406 individuals after exclusions for lipid lowering medications and diabetes. 52 SNPs were genotyped using the Affymetrix 6.0 array or imputed using HapMap. SNPs were excluded for significant deviation from Hardy-Weinberg Equilibrium expectations (P < 1*10 -5 ), imputation quality < 0.5, or minor allele frequency < 0.01. 112 SNP-phenotype associations were conducted using linear models, with SNP as the predictor, and age, sex, data collection site, smoking status, and ancestry-specific principal components of ancestry as covariates. As SNPs associated with each trait were located on different chromosomes, we assumed independence and Fisher’s test was used to combine P-values for each trait, to trait-specific aggregate Q-values (corrected using false discovery rate, Table 1). Results: Association (Q<.01) or suggestive association (.01<Q<.05) was seen for all but one of the 17 traits in the Caucasian group. In Hispanic-Americans, 14/17 of these traits showed association, or suggestively so. There were no associations in the African-American and Chinese-American groups. Conclusions: Aggregate P-values within a trait suggest that overall, different SNPs associate with lipoprotein measures in different ethnicities. Due to the limitations of ethnicity-specific sample size and allele frequencies, we are currently employing weights in a formal test for empirical differences in strength of association between the ethnicities.