A variational Bayes discrete mixture test for rare variant association

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Rare genetic variants are linked to complex traits, but identifying them is challenging.
• Current aggregate tests often ignore neutral variants, hindering the discovery of specific causal variants.
• Existing methods may lack power when only a fraction of variants in a region are causal.

Purpose of the Study:

• To propose a novel aggregate rare-variant test that explicitly models neutral variants.
• To identify specific rare variants driving associations with complex traits at the gene level.
• To improve statistical power compared to existing popular rare-variant association tests.

Main Methods:

• Developed a novel aggregate rare-variant test incorporating a fraction of neutral variants.
• Employed variational Bayes approximate inference for efficient exome-wide analysis.
• Compared the proposed method against Sequence Kernel Association Test (SKAT), Weighted Sum Statistic (WSS), and Morris and Zeggini (MZ) collapsing method via simulations.

Main Results:

• Simulations demonstrated superior power of the novel method in scenarios with a fraction of causal variants.
• The method successfully identified associations between von Willebrand Factor (VWF) levels and rare missense variants in African Americans.
• Approximately 10% of imputed rare missense variants in VWF were strongly associated with lower VWF levels.

Conclusions:

• The proposed method offers increased power for detecting rare variant associations, especially when neutral variants are present.
• It effectively identifies specific causal rare variants contributing to complex traits.
• The approach is computationally efficient for exome-wide studies and has practical applications in genetic research.