A Multiple-Trait Bayesian Variable Selection Regression Method for Integrating Phenotypic Causal Networks in...

Area of Science:

• Genetics
• Bioinformatics
• Statistical Genomics

Background:

• Bayesian regression methods with mixture priors are utilized in multi-trait genomic prediction and can be extended to genome-wide association studies (GWAS).
• Understanding causal structures among traits is crucial for comprehensive multi-trait GWAS to elucidate genotype-phenotype relationships.
• Existing multi-trait GWAS methods may not fully capture complex genetic architectures across multiple traits.

Purpose of the Study:

• To develop a novel GWAS methodology, SEM-Bayesian alphabet, integrating structural equation modeling (SEM) with multi-trait Bayesian regression.
• To incorporate underlying causal structures among traits into GWAS for a more thorough genotype-phenotype mapping.
• To provide a method that analyzes direct, indirect, and overall marker effects for enhanced genetic insights.

Main Methods:

• Development of the SEM-Bayesian alphabet methodology, combining structural equation modeling (SEM) with Bayesian alphabet mixture priors.
• Application of SEM to model causal relationships among multiple traits.
• Performing GWAS by dissecting marker effects into direct, indirect, and overall components.

Main Results:

• SEM-Bayesian alphabet offers a more comprehensive genotype-phenotype understanding compared to traditional multi-trait GWAS.
• The method demonstrated superior performance in GWAS analyses on both simulated and real datasets.
• The study validates the effectiveness of incorporating trait causal structures in multi-trait GWAS.

Conclusions:

• SEM-Bayesian alphabet provides a robust framework for multi-trait GWAS by integrating causal structures.
• This approach enhances the ability to map complex genotype-phenotype relationships.
• The open-source software JWAS facilitates the implementation of these advanced GWAS analyses.