Mining for gene-environment and gene-gene interactions: parametric and non-parametric tests for detecting variance...

Area of Science:

• Genetics
• Statistical Genomics
• Bioinformatics

Background:

• Variance quantitative trait loci (vQTL) detection aids in discovering gene-environment (GxE) and gene-gene (GxG) interactions.
• Identifying vQTLs can streamline GxE and GxG analyses by reducing the number of statistical tests and the associated multiple-testing burden.

Purpose of the Study:

• To conduct a comprehensive head-to-head comparison of various vQTL detection methods.
• To evaluate methods based on false positive rates (FPRs), statistical power, and computational efficiency.

Main Methods:

• Compared three parametric (Deviation Regression Model - DRM, Brown-Forsythe - BF, Double Generalized Linear Model - DGLM) and two non-parametric (Kruskal-Wallis - KW, Quantile Integral Linear Model - QUAIL) vQTL tests using simulation studies.
• Applied the Kruskal-Wallis test (KW) to analyze four lipid traits in the Taiwan Biobank dataset.

Main Results:

• The Deviation Regression Model (DRM) and Kruskal-Wallis test (KW) were identified as the most recommended parametric and non-parametric tests, respectively.
• The Quantile Integral Linear Model (QUAIL) maintained appropriate FPR but had lower power and longer computation times.
• The Brown-Forsythe (BF) test showed inflated FPR with low minor allele frequencies, and the Double Generalized Linear Model (DGLM) was unsuitable for non-normally distributed traits.
• The Kruskal-Wallis test (KW) was robust and computationally efficient, identifying 30 vQTLs for four lipid traits.

Conclusions:

• The Kruskal-Wallis test (KW) is a robust and computationally efficient choice for vQTL detection, particularly for non-normally distributed traits.
• The identified vQTLs in lipid traits were enriched for gene-environment (GxE) and gene-gene (GxG) interactions, validating the utility of vQTL analysis.