Abstracts from the Fifteenth Annual Meeting of the International Genetic Epidemiology Society. St. Petersburg, Florida November 16–17, 2006

Abstract

The evolutionary Malecot model and composite likelihood (CL) have been successfully applied to candidate gene association analysis. To extend the method to genomewide scans, chromosomes were divided into regions based on linkage disequilibrium (LD) maps in LD units (LDUs). A minimum of 10 LDUs and 30 SNPs were assumed for a region individually analyzed. The expected association was predicted by the Malecot model, which is a function of the distance between the disease causing variant and the marker. CL combining associations at multiple loci was maximized to estimate the location of the disease variant. Statistical tests for association were through contrasting hierarchical models. Significance levels were assigned empirically through simulation under the null hypothesis with no association. Starting with a case-control sample, we simulated the case/control status based on genotypes of a randomly chosen SNP taken as causal for a region. Results showed that on average, the estimated point locations for the causal SNPs were only 23 to 30 kb apart from the true locations in these data with SNP density of one per 7.5 kb. Both power and location accuracy depended on the LD between the causal SNP and the nearest surrounding markers, as well as the similarity of minor allele frequencies between them. Also, LD maps provided higher power and location accuracy than kb maps. Our method is both practical and efficient for genome-wide LD scans.