Statistical Methods for Aggregating Trans-Ancestry Genome Wide Association Summary Statistics & Its Applications to Nicotine & Alcohol Addiction Phenotypes

Wang, Xingyan

Statistical Methods for Aggregating Trans-Ancestry Genome Wide Association Summary Statistics & Its Applications to Nicotine & Alcohol Addiction Phenotypes

Restricted

Author:: Wang, Xingyan
Graduate Program:: Biostatistics
Degree:: Doctor of Philosophy
Document Type:: Dissertation
Date of Defense:: October 06, 2022
Committee Members:: Dajiang Liu, Chair & Dissertation Advisor
Jennifer Nyland, Outside Unit Member
Vernon Chinchilli, Major Field Member
Joshua Muscat, Outside Field Member
Rongling Wu, Major Field Member
Arthur Berg, Professor in Charge/Director of Graduate Studies
Keywords:: Trans-Ancestry
Meta-Analysis
GWAS
Smoking and Drinking addiction
Admixed Population
Fine-mapping
TWAS
Gene Set Enrichment Analysis
Abstract:: Large-scale biobank data is available due to the decreasing cost of sequencing in the past decades. This also allows studies to generate summary association statistics at the cohort level that can be shared with researchers while protecting study participants’ privacy. Moreover, by aggregating effects across different studies, we can also increase the power of detecting novel genetic associations between variants and phenotypes of interest. Genome-wide association studies (GWAS) also start to include samples of non-European ancestries. Genetic effects can differ between ancestries, which demands meta-analysis methods to better account for the genetic effect heterogeneities in trans-ancestry meta-analysis. In order to consider the trans-ancestry characteristic, we propose (1) a trans-ancestry meta-analysis method that uses ancestry effects derived from allele frequency to increase power; (2) a trans-ancestry fine-mapping method that can narrow down potential causal variants via statistic methods and quantify loci ancestry heterogeneity; (3) a local ancestry proportion estimation method based on summary statistics for the admixed cohort. We apply our methods to the GWAS and Sequencing Consortium of Alcohol and Nicotine use (GSCAN) study with four smoking-related traits and one drinking-related trait. We show that our methods outperform other methods through simulations and applied data analysis. We further extend our methods to investigate the cross trait genetic architecture, which partitions genetic effects into a component invariant between ancestry, components that vary with ancestry, and a component that varies independently from ancestry. We show that the genetic effect component invariant across ancestries shows the strongest cross-trait genetic correlations, demonstrating that pervasive pleiotropic effects are more likely to be shared across ancestries. We also apply the trans-ancestry TWAS method with GTEx data to discover novel associations at the gene level.

Tools