Statistical analysis in genomic studies

Next-generation sequencing (NGS) technologies reveal unprecedented insights about genome, transcriptome, and epigenome. However, existing quantification and statistical methods are not well prepared for the coming deluge of NGS data. In this dissertation, we propose to develop powerful new statistical methods in three aspects. First, we propose a Hidden Markov Model (HMM) in Bayesian framework to quantify methylation levels at base-pair resolution by NGS. Second, in the context of exome-based studies, we develop a general simulation framework that distributes total genetic effects hierarchically into pathways, genes, and individual variants, allowing the extensive evaluation of existing pathway-based methods. Finally, we develop a new hypothesis testing method for group selection in penalized regression. The proposed method naturally applies to gene or pathway level association analysis for genome-wide data. The results of this dissertation will facilitate future genomic studies.