Summix: A Method for Detecting and Adjusting for Population Structure in Genetic Summary Data

• Activity Number: 415 - Recent advancements in the analysis of large-scale GWAS
• Type: Contributed
• Date/Time: Thursday, August 12, 2021 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Genomics and Genetics
• Author(s): Ian S. Arriaga-MacKenzie* and Gregory Matesi and Samuel Chen and Alexandria Ronco and Katie M. Marker and Jordan R. Hall and Ryan Scherenberg and Mobin Khajeh-Sharafabadi and Yinfei Wu and Christopher R Gignoux and Megan Null and Audrey E Hendricks
• Companies: University of Colorado Denver and University of Colorado Denver and University of Colorado Denver and University of Colorado Denver and University of Colorado Anschutz Medical Campus and University of Colorado Denver and University of Colorado Denver and University of Colorado Denver and University of Colorado Denver and University of Colorado Anschutz Medical Campus and The College of Idaho and University of Colorado Denver
• Keywords: genetics; population structure; sequential quadratic programming; genotype; summary

Abstract

Online aggregation of genetic sequencing data, and the publicly available data produced, are invaluable tools in research and the clinic. These databases have numerous applications including prioritizing causal variants and leveraging common controls. However, summarizing individual-level genotype data can mask population structure, resulting in increased potential for confounding and reduced power. This limits the utility of these databases, especially for understudied and ancestrally diverse populations.

We present Summix, a method to deconvolute ancestry and provide ancestry-adjusted allele frequencies from summary data. Using a continental reference panel, we show our method is accurate and precise to within 0.1% for all simulation scenarios. We apply our method to the Genome Aggregation Database (gnomAD) to estimate ancestry and adjust allele frequencies within known heterogeneous ancestry groups, such as African/African-American (~84% AFR, ~14% EUR) and American/Latinx (~4% AFR, ~5% EAS, ~43% EUR, ~46% IAM).

Summix efficiently runs in seconds, and holds the potential to increase the utility and equity of summary genetic data.