Bayesian detection of causal rare variants under posterior consistency

Area of Science:

• Genetics
• Statistical Genetics
• Genomic Association Studies

Background:

• Identifying causal rare variants for complex traits is a major challenge in genome-wide association studies.
• Current methods often focus on global association, not consistently identifying specific causal variants, especially in small sample sizes with many variants (small-n-large-P).

Purpose of the Study:

• To develop a novel Bayesian method, the Bayesian Rare Variant Detector (BRVD), for robust identification of causal rare variants.
• To simultaneously perform global association tests and detect specific causal variants driving the association.

Main Methods:

• Developed the Bayesian Rare Variant Detector (BRVD) method.
• Incorporated appropriate prior distributions to ensure consistent causal variant identification in small-n-large-P situations.
• Compared BRVD performance against existing methods like C-MVT, WSS, RARECOVER, SKAT, and BRI.

Main Results:

• BRVD demonstrated superior power for global association testing compared to all evaluated methods.
• BRVD showed higher power in identifying causal rare variants than the Bayesian risk index method.
• Applied BRVD to Early-Onset Myocardial Infarction (EOMI) exome sequence data, identifying known causal rare variants.

Conclusions:

• BRVD is a powerful and consistent tool for detecting causal rare variants in complex traits.
• The method effectively addresses both global association and causal variant identification challenges.
• BRVD shows promise for application in real-world genetic datasets like EOMI.