Statistical testing of shared genetic control for potentially related traits

Area of Science:

• Genetics and Bioinformatics
• Statistical Genetics
• Genomic Epidemiology

Background:

• Integrating genome-wide association study (GWAS) data for related traits can help disentangle shared genetic architectures within associated regions.
• Formal statistical colocalization testing is crucial for identifying shared genetic signals but relies on appropriate single nucleotide polymorphism (SNP) selection.

Purpose of the Study:

• To evaluate the impact of SNP selection methods on type 1 error rates in statistical colocalization.
• To propose and validate robust methods for SNP selection to control type 1 error rates.
• To apply refined colocalization methods to investigate shared genetic factors between Graves' disease and Hashimoto's thyroiditis.

Main Methods:

• Simulations were used to assess the performance of different SNP selection strategies in colocalization analyses.
• Methods evaluated included testing principal components and Bayesian model averaging to control type 1 error rates.
• Colocalization analysis was applied to genetic association data from Graves' disease and Hashimoto's thyroiditis.

Main Results:

• Published SNP selection methods can lead to substantially inflated type 1 error rates.
• Testing informative principal components or using Bayesian model averaging effectively controls type 1 error rates.
• Analysis revealed a common genetic signature in seven regions for Graves' and Hashimoto's thyroiditis, suggesting genuine shared associations.

Conclusions:

• Robust SNP selection is critical for accurate statistical colocalization and avoiding inflated type 1 errors.
• The proposed methods enhance the reliability of colocalization analyses, facilitating the discovery of shared genetic architectures.
• Colocalization analysis holds significant potential for identifying shared genetic signatures across related diseases and pinpointing causal genes and tissues.