Statistical fine-mapping and summary statistics imputation in genome-wide association studies

Abstract

In genome-wide association studies (GWASs), a large number of trait-associated single-nucleotide polymorphisms (SNPs) have been detected. Among these associations, not all SNPs are the causal ones due to the correlation between SNPs. Many statistical fine-mapping methods have been proposed to identify SNPs that mechanistically affect a disease/trait. These existing methods need information of all SNPs for the fine-mapping purpose. Before the whole-genome sequencing technique is widely used in genome-wide association studies to genotype every SNP in the genome, we need to rely on imputation methods to obtain genotypes or summary statistics of all SNPs. However, existing imputation methods assume the effect sizes are all zero in their calculations. This unrealistic assumption has introduced some error to the imputation results, and has therefore affected the accuracy of fine-mapping. In this thesis, we propose a novel method to carry out fine-mapping and imputation jointly to avoid the flawed assumption in current imputation methods. Experiments with simulation data and real data illustrate that our new method outperforms traditional fine-mapping methods in both accuracy and speed. For imputation, our method also shows slightly improvement in accuracy compared with traditional methods.

Date of Award	2019
Original language	English
Awarding Institution	The Hong Kong University of Science and Technology