Regularized multi-trait multi-locus linear mixed models for genome-wide association studies and genomic selection in...

Area of Science:

• Genomics
• Quantitative Genetics
• Plant Breeding

Background:

• Multi-trait genome-wide association studies (GWAS) and genomic selection (GS) are crucial for understanding complex traits.
• Existing multi-trait linear mixed models struggle with high-dimensional genotype data and computational burden.
• Previous regularization methods for multi-trait linear models overlooked population structure and familial relatedness.

Purpose of the Study:

• To develop novel regularized multi-trait linear mixed models for genomics.
• To create scalable estimation approaches for high-dimensional genotype and multi-trait data.
• To improve accuracy in genomic selection and identify marker-trait associations.

Main Methods:

• Proposed a new class of regularized multivariate linear mixed models.
• Developed scalable estimation methods for high-dimensional data.
• Applied methods to maize and sorghum diversity panels.

Main Results:

• Demonstrated high prediction accuracy in genomic selection (GS).
• Successfully identified relevant marker-trait associations.
• Validated the effectiveness of the proposed models on real-world plant datasets.

Conclusions:

• The developed models are effective for both GWAS and GS.
• These regularized multivariate linear mixed models offer advancements for plant biology and crop breeding.
• Facilitates research in agronomy by enabling better genetic insights and breeding strategies.