Nonparametric linkage analysis using person-specific covariates.

Abstract

Linkage analysis provides an important tool for mapping genes for complex disease. However its usefulness has been limited by inadequate marker density, inadequate sample sizes and the possibility that different genes account for different subtypes of the disease (phenotypic heterogeneity). The first two limitations can be addressed by high-density single nucleotide polymorphism (SNP) genotyping and the pooling of large sets of multiple-case families. Phenotypic heterogeneity can be addressed by analyses that weigh the contributions of affected family members according to characteristics of their disease phenotypes. Here we introduce a method for including such person-specific weights in nonparametric linkage analysis. We show with simulations that such weighting can provide stronger linkage signals when a causal polymorphism affects some manifestations of the disease more than others. We applied the method to prostate cancer linkage data in a region on chromosome 19p, and obtained higher lod scores by assigning weights of one to men with early-onset aggressive cancers, weights of zero to those with late-onset nonaggressive cancers, and intermediate weights to all other affected men. We have developed a modified version of GENEHUNTER that allows inclusion of person-specific weights in the nonparametric analyses. This program is freely available at http://med.stanford.edu/epidemiology/statisticalSoftware/weightedKAC.