Right-hand-side updating for fast computing of genomic breeding values

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Genomic prediction models face increasing computational demands due to growing numbers of single nucleotide polymorphisms (SNPs) and larger training datasets.
• Efficient genomic prediction is crucial for accurate genetic evaluations.

Purpose of the Study:

• To develop and evaluate novel algorithms for genomic prediction that enhance computational efficiency.
• To reduce the computing time and memory (RAM) required for genomic prediction models.

Main Methods:

• Two alternative algorithms, "improved residual updating" and "right-hand-side updating" (RHS-updating), were developed to replace the original residual updating algorithm.
• The algorithms leverage the three-value nature of SNP genotypes and extend updating across multiple SNPs.
• Implementation and testing were performed using a Bayesian stochastic search variable selection model.

Main Results:

• The improved residual updating algorithm decreased CPU time by 35.3–43.3% with no change in memory requirements.
• The RHS-updating algorithm achieved substantial reductions: 74.5–93.0% in CPU time and 13.1–66.4% in memory requirements compared to the original algorithm.

Conclusions:

• The RHS-updating algorithm offers a significant improvement in computational efficiency for genomic prediction.
• This algorithm is a valuable alternative for reducing both computing time and memory usage in various genomic prediction models.