Design and Association Methods for Next-generation Sequencing Studies for Quantitative Traits.

GWAS;Next-generation Sequencing;Association Methods;Study Design;Meta-analysis;MCMC;Pedigree;Family Sample;Sample Selection;Quantitative Traits;Association Studies;Statistics and Numeric Data;Science;Biostatistics

Feng, ShuangSong, Peter Xuekun ;

University of Michigan

关键词: GWAS;    Next-generation Sequencing;    Association Methods;    Study Design;    Meta-analysis;    MCMC;    Pedigree;    Family Sample;    Sample Selection;    Quantitative Traits;    Association Studies;    Statistics and Numeric Data;    Science;    Biostatistics;   

Others  :  https://deepblue.lib.umich.edu/bitstream/handle/2027.42/113521/sfengsph_1.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

Advances in exome sequencing and the development of exome genotyping arrays are enabling explorations of association between rare coding variants and complex traits using sequencing-based GWAS. However, the cost of sequencing remains high, optimal study design for sequencing-based association studies is an open question, powerful association methods and software to detect trait-associated rare and low-frequency variants are in great need. Containing 5% of information in human genome, chromosome X analysis has been largely neglected in routine GWAS analysis. In this dissertation, I focus on three topics:First, I describe a computationally efficient approach to re-construct gene-level association test statistics from single-variant summary statistics and their covariance matrices for single studies and meta-analyses. By simulation and real data examples, I evaluate our methods under the null, investigate scenarios when family samples have larger power than population samples, compare power of different types of gene-level tests under various trait-generating models, and demonstrate the usage of our methods and the C++ software, RAREMETAL, by meta-analyzing SardiNIA and HUNT data on lipids levels.Second, I describe a variance component approach and a series of gene-level tests for X-linked rare variants analysis. By simulations, I demonstrate that our methods are well controlled under the null. I evaluate power to detect an autosomal or X-linked gene of same effect size, and investigate the effect of sex ratio in a sample to power of detecting an X-linked gene. Finally I demonstrate usage of our method and the C++ software by analyzing various quantitative traits measured in the SardiNIA study and report detected X-linked variants and genes.Third, I describe a novel likelihood-based approach and the C++ software, RAREFY, to prioritize samples that are more likely to be carriers of trait-associated variants in a sample, with limited budget. I first describe the statistical method for small pedigrees and then describe an MCMC approach to make our method computationally feasible for large pedigrees. By simulations and real data analysis, I compare our approach with other methods in both trait-associated allele discovery power and association power, and demonstrate the usage of our method on pedigrees from the SardiNIA study.