|
Abstract:
|
With case-parent data, joint analysis of multiple markers can be useful. Risk may depend on multiple loci that together code for a dysfunctional protein. Alternatively, the true causative variant may be a SNP that was not typed but is in linkage disequilibrium with markers that were. Studying joint effects of a haplotype and an exposure is challenging, because population structure can produce biased tests. Under a null hypothesis that there is no interaction (multiplicative scale), transmission of a causative haplotype from parents to affected offspring would be independent of exposure. We developed a nonparametric test statistic, referenced to permutations within strata based on sums of unphased parental genotypes. Simulations suggest reasonable power. We apply the method to maternal smoking, the GSTP1 gene, and a birth defect, oral clefting.
|
