Evaluating confounding in rare variant genome wide association studies.

Area of Science:

• Genetics
• Population Genetics
• Statistical Genetics

Background:

• Population-based genetic studies risk confounding from rare variant associations.
• Empirical investigation of this risk is lacking.
• Understanding confounding is crucial for accurate genetic discoveries.

Purpose of the Study:

• To empirically investigate the risk of confounded rare variant associations in population genetic studies.
• To assess the effectiveness of current methods in capturing demography and mitigating bias.
• To explore confounding mechanisms and the utility of family-based analyses.

Main Methods:

• Analysis of 306,991 sequenced exomes from the UK Biobank.
• Evaluation of common and rare variant principal components for capturing recent demography.
• Re-analysis of 155 phenotypes in siblings to assess bias in effect estimates.
• Investigation of confounding from haplotype sharing, spatial structure, assortative mating, and local linkage.

Main Results:

• Recent demography is poorly captured by standard genetic components.
• Accounting for haplotype sharing does not eliminate false-positive rare variant associations with non-heritable traits.
• Bias in effect estimates is higher for non-uniformly distributed traits, indicating pervasive population stratification.
• Assortative mating significantly influences bias in height-related rare variant associations.
• Risk of elevated false discovery rates exists for recent variants in extended families and through local linkage.

Conclusions:

• Complex confounding mechanisms impact rare variant studies.
• Current methods are insufficient for fully capturing demographic effects and preventing bias.
• Family-based approaches offer valuable sensitivity analyses for rare variant association studies.