Friday, June 26 2020
10:00am - 12:00pm
PhD Thesis Presentation
PhD Presentation - Emileigh Willems

Title: Increasing Power of Trans-ethnic Meta-analysis Methods by Enhanced Utilization of Admixed Samples

Committee: Stephanie Santorico (Advisor), Audrey Hendricks (Chair), Erin Austin, Joshua French, Karen Edwards

Analyzing human genetic data provides a better understanding of the human genome and how it relates to health outcomes. This has led to vast improvements in precision medicine, which better tailors an individual's medical treatment to their genetic information. The risk and burden of diseases and health outcomes often varies between ancestries, due to genetic differences or environmental factors, so care must be taken to properly model genetic ancestry in all genetic analyses to avoid spurious results. Further, the size and number of large ancestrally diverse human genetic data sets has greatly increased in the last decade, opening a need to develop statistical methods that properly analyze such resources. Recent advancement and discovery from analyzing genetic data from diverse ancestries, particularly, analyzing samples from admixed ancestry (e.g., African American and Latinx populations), have produced powerful genome-wide association studies due to unique features within admixed samples.

Meta-analysis or mega-analysis methods can be used to analyze large samples of human genetic data, and recent developments in both classes of statistical models have been motivated by properly modeling the large amount of ancestral diversity present in humankind. However, these advancements do not typically prioritize modeling admixed samples at the meta-analysis or mega-analysis level. This dissertation seeks to evaluate and improve methods that analyze large samples of genetic data from ancestrally diverse backgrounds and specifically assess if these methods are able to capture the unique information available within admixed samples.

First, an empirical comparison of existing meta-analysis methods on the ancestrally diverse GENNID study was performed and evaluated. This motivated a comprehensive simulation study comparing meta-analysis and mega-analysis methods' ability to properly incorporate admixed samples. The use of different genetic distance metrics is also compared within meta-analysis methods, to assess if these metrics are able to better model the properties of admixed samples. A comprehensive simulation plan is presented, with all details of the simulation framework. The simulation results give insight into existing statistical methods' behavior when including admixed samples and evaluate topics that should be considered by researchers when designing their own analyses for large ancestrally diverse samples. The use of different genetic distance metrics are evaluated in meta-analysis methods and show need for further improvement to properly model admixed samples at the meta-analysis level.
Speaker:Emileigh Willems
Location:Zoom meeting

Download as iCalendar