Comparative study of statistical methods for detecting association with rare variants in exome-resequencing data

Area of Science:

• Genetics
• Statistical Genetics
• Genomic Association Studies

Background:

• Genome-wide association studies (GWAS) typically assume common disease/common variant (CDCV) or common disease/rare variant (CDRV) models.
• Classical single-marker tests are effective for CDCV but less powerful for CDRV.
• New methods are needed to collectively detect associations with multiple rare variants within genes.

Purpose of the Study:

• To compare the performance of novel rare variant association methods using Genetic Analysis Workshop 17 data.
• To evaluate these methods on both unrelated and family-based datasets for a quantitative trait (Q1).
• To assess the power and type I error control of different rare variant detection approaches.

Main Methods:

• Applied several recently developed rare variant association methods to the Genetic Analysis Workshop 17 dataset.
• Utilized unrelated individuals and family data for association testing.
• Conducted analyses across 200 replicates for the quantitative trait Q1.

Main Results:

• Collapsing methods demonstrated potential for detecting associations with rare variants.
• Power to detect associations was generally low in the analyzed dataset.
• Type I error rates were not consistently controlled across methods, varying by gene.
• Collapsing and single-locus methods may exhibit differential susceptibility to population stratification.

Conclusions:

• Collapsing methods represent a promising avenue for rare variant association studies.
• Further research is required to address type I error control and validate power comparisons.
• Investigating the impact of population stratification on different association approaches warrants further attention.