Methods to Improve the use of Common Controls in Sequencing Studies

Event Information

 

Who: Dr. Audrey Hendricks

What: 2019 Spring Seminar Series

When: Friday, May 3rd at 12:00 PM

Where: SI 2001

Dr. Audrey Hendricks photoDr. Audrey Hendricks
Assistant Professor
Mathematical and Statistical Sciences
Colorado School of Public Health

"Methods to Improve the use of Common Controls in Sequencing Studies"

Uncovering the functional mechanism of genetic associations across different variants, environmental and genetic backgrounds, tissues, and phenotypes is challenging and requires a large amount of resources. Additionally, gathering and completing adequately sized and powered studies of complex diseases to identify novel genetic associations is expensive. As such, wise allocation of limited resources is essential. Large genetic resources such as the Genome Aggregation Database (gnomAD, ~140,000 sequenced samples), NHLBI’s Trans-Omics for Precision Medicine (TOPMed) BRAVO interface (>60,000 genomes), and NHGRI’s Genome Sequencing Program (GSP, ~20,000 genomes for use as controls) have the potential to be used as controls in studies of complex diseases. Differences between internal data and external data will exist creating a large potential for confounding and an increase in type I and type II error. Robust methods are needed to appropriately use these data. Additionally, potential common control samples are often heterogenous with respect to ancestry and may contain individuals with common diseases (e.g. Type 2 Diabetes). This heterogeneity can further confound or decrease power to detect association with case status. Here, I present two approaches to improve the use of common controls from sequencing studies. First, Proxy External Controls Association Test (ProxECAT) tests for association of a genetic region with case status while controlling for differences in sequencing data generation. Second, identifying hidden ancestries estimates the proportion of global ancestry group within summary genetic data. Both methods utilize publicly available frequency level data enabling broad and efficient use of external data increasing our ability to detect genetic associations while adequately controlling for confounding.

Everyone is welcome to the seminar. If you would like to meet with the speaker, please email Dr. Greg Ragland at gregory.ragland@ucdenver.edu.