Joint regression analysis of multiple traits based on genetic relationships

Area of Science:

• Genomics
• Statistical Genetics
• Bioinformatics

Background:

• Polygenic scores (PGSs) are crucial for predicting genetic risk and understanding genetic architectures.
• Current methods for analyzing PGSs often lack the ability to infer trait clusters or have limited predictive power.
• Multivariate genome-wide association studies (GWAS) present challenges in handling correlated traits.

Purpose of the Study:

• To introduce geneJAM, a novel clustering and estimation method for inferring genetic relationships among multiple traits in multivariate GWAS.
• To leverage PGSs for identifying groups of traits that share underlying genetic characteristics.
• To enhance the predictive capabilities and analytical efficiency in genetic studies of complex traits.

Main Methods:

• Utilizes graphical lasso to estimate a sparse covariance matrix of PGSs, revealing trait clusters.
• Employs identified clusters to structure the error covariance matrix in a generalized least squares (GLS) model.
• Applies feasible GLS for estimating linear regression models with correlated residuals.

Main Results:

• Successfully identifies clusters of traits with shared genetic characteristics in multivariate GWAS.
• Demonstrates increased precision, statistical power, and computational efficiency compared to existing methods.
• Validated on simulated data and a heterogeneous stock mouse dataset, showcasing practical utility.

Conclusions:

• geneJAM provides a robust framework for analyzing complex genetic relationships among multiple traits using PGSs.
• The method facilitates the development of PGS research by enabling the discovery of biologically meaningful trait clusters.
• geneJAM enhances the analysis of multivariate GWAS, offering improved accuracy and efficiency for biological studies.