Divided-and-combined omnibus test for genetic association analysis with high-dimensional data.

Area of Science:

• Genomics and Bioinformatics
• Statistical Genetics
• High-Dimensional Data Analysis

Background:

• Biologic technology generates vast high-dimensional genetic and genomic data.
• Analyzing associations between multiple phenotypes and variants is crucial but challenging.
• Traditional methods suffer power loss due to multiple testing adjustments and ignoring phenotype correlations.

Purpose of the Study:

• To address the power limitations of similarity-based methods in high-dimensional genetic association studies.
• To develop a novel statistical test that accounts for correlations among multiple phenotypes.
• To improve the robustness and power of genetic association analysis for complex datasets.

Main Methods:

• Proposed a divided-and-combined omnibus test utilizing a divided-and-combined strategy.
• Signals are divided into groups based on phenotypic similarity matrix eigenvalues.
• Analysis results are combined using the Cauchy-combined method to generate a final statistic.

Main Results:

• The proposed test demonstrated significantly higher power and robustness compared to the original similarity-based method.
• Power increases of over 0.6 were observed in certain scenarios.
• The method was validated through extensive simulations and application to pig genetic data.

Conclusions:

• The divided-and-combined omnibus test effectively handles drawbacks of existing similarity-based methods for high-dimensional data.
• This approach facilitates more powerful genetic association analysis, particularly when dealing with correlated phenotypes.
• The test is broadly applicable to association analyses between any two multivariate variables.