A variance components factor model for genetic association studies: a Bayesian analysis.

Area of Science:

• Genetics and Bioinformatics
• Statistical Modeling
• Complex Disease Research

Background:

• Gene-trait associations for complex diseases often involve multiple, correlated traits.
• These traits can manifest as lower-dimensional latent factors or disease syndromes.
• Univariate models applied separately to each trait may miss complex interrelationships.

Purpose of the Study:

• To introduce and illustrate the Variance Components Factor (VCF) model for analyzing multiple gene-trait associations.
• To model associations between multiple traits, genotype groups, and patient-level covariates.
• To provide a Bayesian implementation of the VCF model for clinical decision-making.

Main Methods:

• Application of the Variance Components Factor (VCF) model within a Bayesian framework.
• Implementation using WinBUGS software for straightforward analysis.
• Illustration with simulated data and a real-world example involving apolipoprotein-E gene polymorphisms and lipid measurements.

Main Results:

• The VCF model successfully characterizes multiple trait manifestations across genotype groups, even with unobserved group assignments.
• The model allows investigation of covariate by genotype group interactions influencing trait variability.
• Simulations confirm the model's ability to capture complex trait patterns.

Conclusions:

• The VCF model offers a flexible and desirable alternative to univariate models for analyzing multiple correlated traits in genetic studies.
• It enhances understanding of latent factors underlying complex diseases and their genetic basis.
• The Bayesian approach facilitates the incorporation of prior knowledge, improving model robustness.