Study design issues in the analysis of complex genetic traits

Abstract

Many common diseases that potentially have a large public health impact, such heart disease, cancer, and diabetes, are known or thought to have a genetic component to risk. These traits are often complex with multiple contributing genetic and non-genetic factors. The identification of genetic risk factors through genetic mapping and cloning studies may aid our understanding of the underlying mechanism of disease. Unfortunately large sample sizes are often required for linkage analysis of complex traits, resulting in a need to identify cost-effective strategies. The use of power and sample size considerations as the criteria for evaluating methods may grossly underestimate the utility of some designs. Additional factors can be incorporated into the analysis by considering the overall cost of a study including the pedigree collection, marker genotyping, and statistical analysis. Using analytical and simulation methods, we evaluate three study design issues relevant to cost-effective linkage analysis of complex traits.First, we demonstrate that pedigrees of the same size and structure that differ in the number of affected and unaffected individuals also differ in the probability of segregating a particular trait locus, which affects the power to detect linkage at that locus. Although the optimal pedigrees for detecting linkage to a particular trait locus are highly dependent on the trait model, we identify pedigree selection schemes that are cost-effective for a variety of underlying trait models. Second, we identify characteristics of single nucleotide polymorphisms (SNPs) for a cost-effective genome screen compared to the current microsatellite-based technology. Issues that are addressed in comparing the markers include the information content for clustered SNPs in the presence or absence of linkage disequilibrium, the marker spacing, and the map accuracy. Finally, we describe a method called 'downcoding' to reduce the computational burden associated with currently available likelihood-based methods that are potentially more powerful for detecting linkage than alternative methods. It is often not possible to use these methods without downcoding because of limitations in the available computational resources. Through consideration of the three topics described above, we are able to identify cost-effective strategies for the analysis of complex genetic traits.