Multiple-trait genome-wide association study based on principal component analysis for residual covariance matrix

Area of Science:

• Genetics
• Statistical genetics
• Bioinformatics

Background:

• Multivariate analysis in genome-wide association studies (GWAS) presents implementation challenges.
• Principal component analysis (PCA) is often used to simplify multiple traits but can lead to inaccurate parameter estimates.
• Existing methods may yield spurious linkage results due to discrepancies between univariate and joint analyses.

Purpose of the Study:

• To propose a novel PCA-based method for multivariate trait analysis in GWAS.
• To enhance the accuracy of parameter estimates compared to traditional PCA methods.
• To improve the efficiency and reduce computational costs in genetic trait mapping.

Main Methods:

• Performing PCA on the residual covariance matrix instead of the phenotypic covariance matrix.
• Transforming multiple traits into pseudo principal components for separate analysis.
• Utilizing a fast least absolute shrinkage and selection operator (LASSO) for sparse genetic model estimation.

Main Results:

• The proposed method yields parameter estimates equivalent to joint multivariate analysis.
• Statistical and computational efficiencies were demonstrated through extensive simulations.
• The method effectively reduces computational costs associated with complex genetic models.

Conclusions:

• PCA on residual covariance matrices provides a statistically sound and computationally efficient approach for GWAS.
• This method overcomes limitations of traditional PCA in multivariate trait analysis.
• The approach is applicable to complex traits, as shown in a beef cattle GWAS.