Dynamic Bayesian testing of sets of variants in complex diseases

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Genome-wide association studies (GWAS) increasingly focus on rare genetic variants for complex human diseases.
• Current methods often test predefined variant sets separately, limiting power and potentially excluding common variants.
• Existing approaches struggle with accurate variant selection and can be confounded by linkage disequilibrium.

Purpose of the Study:

• To develop a novel Bayesian method for simultaneous, genome-wide testing of all genetic variants.
• To improve power and flexibility in identifying disease-associated variants across the full frequency spectrum.
• To account for variant set correlations and distinguish direct from indirect associations.

Main Methods:

• A Bayesian framework for simultaneous testing of all genome-wide variants.
• Flexible, dynamic grouping and joint testing of variants.
• Accounting for correlations among variant sets to report direct associations only.
• Inclusion of categorical, discrete, and continuous covariates.

Main Results:

• The novel method demonstrates comparable power to existing methods for very rare variants in small sets.
• Significantly outperforms existing methods when including moderately rare or common variants, or testing large variant collections.
• Successfully applied to a whole-genome resequencing study of type 1 diabetes, showcasing its utility and power.

Conclusions:

• The proposed Bayesian method offers enhanced power and flexibility for genome-wide association studies.
• It effectively identifies multiple disease variants with high resolution by simultaneously testing all variants.
• This approach represents a significant advancement over existing rare variant mapping methods, particularly for complex diseases.