Statistical methods for linkage disequilibrium mapping of disease susceptibility genes in candidate regions

Within tightly linked regions of the genome it is often inefficient to genotype and test all polymorphisms for an association with disease since the high degree of linkage disequilibrium causes high levels of redundancy. It has long been suggested that an informative subset of these polymorphisms should be selected as markers for analysis. Within the approach of this thesis, a causal polymorphism takes on the role of a latent variable, generating indirect association between phenotype and marker loci which are in linkage disequilibrium with the causal variant. Locally efficient test statistics are derived using arguments based upon incomplete data likelihood theory and their potential power is investigated. The central chapter sets up a formal mathematical framework on which all subsequent analysis is based, and focusing upon the simple multiplicative model, it investigates how the choice of markers and the degree of haplotype information included may affect the power of the proposed test statistic. The multiplicative disease model is then extended to the dominance model and a test statistic that allows for an indirect dominance effect is derived. The loss or gain in power that may occur in the absence or presence of an underlying dominance effect is examined. The thesis also considers the case of gene-gene interactions and addresses the question: when it is more powerful to test for numerous pairs of possibly interacting loci rather than testing their main effects singularly? This problem is considered in the case of known causal variants and then extended under the latent variable model, deriving a new test statistic in the process. The loss in power caused by allowing for unknown population structure is studied, as is the extent of power lost by analysing pooled genotype data.EThOS - Electronic Theses Online ServiceGBUnited Kingdo