Invited review: efficient computation strategies in genomic selection.

Area of Science:

• Animal Genetics
• Computational Biology
• Quantitative Genetics

Background:

• Genomic selection (GS) is crucial for accelerating genetic gain in animal breeding.
• Accurate genomic prediction requires efficient computational methods for large datasets.
• Existing methods face challenges with scalability and computational cost.

Purpose of the Study:

• To review and evaluate computational methods for genomic selection in animal breeding.
• To identify efficient algorithms for genomic prediction, parameter estimation, and genome-wide association studies (GWAS).
• To assess the impact of computational advancements on the feasibility of GS in large populations.

Main Methods:

• Evaluation of commonly used models: SNP BLUP, GBLUP, and ssGBLUP.
• Analysis of solving methods: Cholesky decomposition, Gauss-Seidel (GS), and preconditioned conjugate gradient (PCG), including iteration on data.
• Assessment of specialized matrix inversion techniques (e.g., APY) and sparse solvers (e.g., YAMS).

Main Results:

• Iterative methods (GS, PCG) with iteration on data are efficient for SNP BLUP and general solutions, respectively.
• The APY inverse enables GBLUP and ssGBLUP for large populations.
• The YAMS package significantly improves performance for sparse systems with dense blocks, allowing GREML on large populations.

Conclusions:

• Genomic selection is computationally maturing, with advanced methods enhancing scalability.
• Efficient algorithms and specialized solvers are key to applying GS in large-scale animal breeding programs.
• The reviewed methods facilitate accurate genomic prediction and GWAS, driving genetic progress.