Analyze multivariate phenotypes in genetic association studies by combining univariate association tests

Area of Science:

• Genetics
• Biostatistics
• Computational Biology

Background:

• Multivariate phenotypes are common in genome-wide association studies (GWAS).
• Exploiting multivariate phenotypes for detecting pleiotropic genetic effects remains challenging.
• Limited methods exist for mixed quantitative and qualitative measures in multivariate phenotypes.

Purpose of the Study:

• To evaluate existing methods for combining univariate test statistics in GWAS.
• To propose and assess novel extensions for analyzing multivariate phenotypes.
• To enhance the power of detecting genetic variants with pleiotropic effects.

Main Methods:

• Evaluation of O'Brien's and Wei and Johnson's combined test statistics approach.
• Development and simulation of two extended multivariate phenotype analysis methods.
• Application to genome-wide association studies (GWAS) data including continuous, categorical, and survival phenotypes.

Main Results:

• All tested methods demonstrated valid type I error rates.
• Proposed extensions showed improved power over O'Brien's method with heterogeneous means.
• Methods offered increased power compared to individual univariate testing in certain scenarios.

Conclusions:

• The developed methods effectively analyze multivariate phenotypes in GWAS.
• These approaches increase the potential for identifying pleiotropic genetic variants.
• The methods are applicable to diverse sample types and phenotype components.