Comparison of statistical tests for association between rare variants and binary traits

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Genome-wide association studies (GWAS) identify common variants but leave much genetic contribution unexplained.
• Rare variants are a potential source of missing heritability for complex traits and diseases.
• Statistical analysis of rare variants is challenging due to their low frequency and potentially heterogeneous effects.

Purpose of the Study:

• To extend a previously developed method for analyzing quantitative traits to accommodate binary traits with covariates.
• To evaluate the performance of the extended method against existing statistical approaches for rare variant analysis.
• To propose a computationally efficient two-step strategy for genetic association studies involving rare variants.

Main Methods:

• Extension of a statistical method to analyze rare variants in binary traits, incorporating covariates.
• Simulation studies under various causal and covariate impact scenarios.
• Comparison of the proposed method with standard logistic regression, C-alpha, SKAT (Sequentially Weakly Associated regions), and EREC (Efficient Rare-variant Association test).

Main Results:

• Logistic regression performs well when effect heterogeneity is low.
• SKAT and EREC show good performance across scenarios but can be computationally intensive.
• The proposed method and a two-step strategy offer a balance between performance and computational efficiency.

Conclusions:

• A novel statistical method is presented for rare variant analysis in binary traits with covariates.
• A practical two-step approach is recommended: gene selection using faster methods followed by analysis with computationally intensive methods like SKAT/EREC.
• The choice of gene selection method depends on assumptions about effect heterogeneity and covariate impact.