Multilocus association analysis under polygenic models

We develop an analysis method for genome-wide case-control association studies that is based on a polygenic threshold model. For each SNP in a given study, the risk allele is determined as that allele leading to an odds ratio greater than 1. For a given set of SNPs, the number of risk alleles in cases minus that in controls is evaluated and a p-value is obtained for this difference. For SNPs selected in a given order based on some single-locus test statistic, successive sums of these differences over the best 2, 3, etc. SNPs (located anywhere in the genome) and associated p-values are obtained. The smallest such p-value among L SNPs tested is our genome-wide test statistic, for which an empirical significance level is obtained by permutation analysis. Our approach is applied to several disease datasets and shown to furnish significant results even for traits with little evidence of single-locus effects.