Bayesian methods for fine mapping in genetic association studies

Thesis

Over the past several years genetic variation has been the centre of attention for different branches of genetics. The use of genome wide association studies (GWAS) has facilitated the discovery of a plethora of genetic variants which are associated with a vast range of diseases. These GWAS primarily examine a single variant, mostly a Single Nucleotide Polymorphism (SNP), at a time based on genotype data. Although this practice has proven to be powerful, low percentages of heritability prompt a call for the exploration of more complex structures of association.

As diseases are more likely to be caused by several variants, multi-SNP analyses are used to refine the signal leading to the development of fine mapping methodologies. Further biological information is accounted for by studying the diploid nature of the human genome which identifies phenomena, such as compound heterozygosity. We propose a flexible Bayesian fine mapping model for quantitative traits allowing for the analysis of haplotype data under a variety of genotype and haplotype models. A Metropolis-Hastings algorithm is employed to search the model space efficiently. Simulations show competitive performance over existing methods when compound heterozygous effects are present in the data.

An extension of a single cohort framework is meta-analysis where a collection of studies is used to increase confidence and boost power in noticeable associations. Traditional approaches include the fixed effects model where the assumption is that genetic effects are homogeneous amongst cohorts while the random effects model accounts for heterogeneity. We propose a Bayesian meta-analysis approach which implements both fixed and random effects with the addition of a parameter measuring the existence of effect in each cohort. Simulations and application to real samples explore the performance of the method under different conditions and show comparable evidence over established methodologies.

Authors: Frangou, E

• Publication date: 2013
• Type of award: MSc by Research
• Level of award: Masters
• Awarding institution: Oxford University, UK
• Language: English
• Keywords: bayesian; genetics; random effects; mcmc; fine mapping; meta analysis; compound heterozygosity; statistics
• Subjects: Mathematical genetics and bioinformatics (statistics)