BayesRVAT enhances rare-variant association testing through Bayesian aggregation of functional annotations.

Area of Science:

• Genetics
• Computational Biology
• Statistical Genetics

Background:

• Gene-level rare variant association tests (RVATs) are crucial for understanding disease mechanisms and finding therapeutic targets.
• Machine learning advances provide numerous variant pathogenicity scores, but current RVATs have limitations in leveraging these effectively due to rigid models or single annotations.

Purpose of the Study:

• To introduce BayesRVAT, a novel Bayesian rare variant association test designed to overcome limitations of existing methods.
• To enable joint modeling of multiple annotations for improved RVAT performance.

Main Methods:

• BayesRVAT employs a Bayesian framework to jointly model diverse genetic annotations.
• It specifies priors on annotation effects and estimates gene- and trait-specific posterior burden scores.
• The method flexibly captures various rare-variant architectures.

Main Results:

• Simulations demonstrate that BayesRVAT offers improved statistical power while maintaining calibration.
• Analysis of UK Biobank data revealed 10.2% more blood-trait associations compared to existing methods.
• Novel gene-disease links were identified, including PRPH2 with retinal disease.

Conclusions:

• BayesRVAT provides a flexible and powerful approach for rare variant association testing.
• Integrating BayesRVAT into omnibus frameworks further enhances discovery by capturing complementary signals.
• The method advances the ability to uncover genetic underpinnings of complex traits and diseases.