Strategies and statistical methods for linkage disequilibrium-based mapping of complex traits

• Main Author: Jia, Tianye
• Published: University of Birmingham 2012
• Subjects: 576.58; QR Microbiology
• Online Access: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.549901

Nowadays, there are many statistical methods available for genetic association analyses with data various designs. However, it is usually ignored in these analyses that an analytical method must be appropriate for an experimental design from which data is collected. In addition, association study is a population-based analysis and, thus its inference is highly vulnerable to many population-oriented confounding factors. This thesis starts with a comprehensive survey and comparison of those methods commonly used in the literature of genetic association study in order to obtain insights into the statistical aspects and problem of the methods. On the basis of these reviews, we managed to calculate the optimal trend set for the Armitage’s trend test for different penetrance models with a high level of genetic heterogeneity. We introduced two new strategies to adjust for the population stratification in association analyses. We proposed a maximum likelihood estimation method to adjust for biases in statistical inference of linkage disequilibrium (LD) between pairs of polymorphic loci by using non-random samples. In the process of the analysis, we derived a more sophisticated but robust likelihood-based statistical framework, accounting properly for the non-random nature of case and control samples. Finally, we developed a multi-point likelihood-based statistical approach for a genome-wide search for the genetic variants that contribute to phenotypic variation of complex quantitative traits. We tested these methods through intensive simulation studies and demonstrated their application in analyses with large case and control SNP datasets of the Parkinson’s disease. Despite that we have mainly focused on SNP data scored from microarray techniques, the theory and methodology presented here paved a useful stepping stone approach to the modeling and analysis of data depicting genome structure and function from the new generation sequencing techniques.