Efficient methods to compute genomic predictions.

Area of Science:

• Animal Genetics
• Bioinformatics
• Quantitative Genetics

Background:

• Accurate estimation of breeding values is crucial for genetic improvement in livestock.
• Traditional methods using pedigree information have limitations in capturing genomic information.
• Genomic selection offers potential to increase prediction accuracy by utilizing high-density marker data.

Purpose of the Study:

• To develop and evaluate efficient methods for processing large-scale genomic data.
• To improve the reliability of estimated breeding values (EBVs) using genomic information.
• To estimate thousands of marker effects simultaneously for enhanced genetic evaluations.

Main Methods:

• Development of algorithms and computer programs for genomic data processing.
• Utilized simulated data from 2,967 bulls and 50,000 random markers across 30 chromosomes.
• Employed three equivalent linear model prediction methods: iteration for allele effects, selection index with genomic relationships, and mixed model equations.

Main Results:

• Achieved a reliability of 63% for predicted net merit in young bulls, a substantial increase from 32% using traditional relationship matrices.
• Nonlinear predictions further improved reliability to 66% for young bulls.
• Efficient processing allowed allele frequency estimation in 2 processor days and genomic predictions in <1 day per trait.

Conclusions:

• Developed efficient genomic data processing methods that significantly increase the reliability of estimated breeding values.
• Genomic predictions provide information equivalent to approximately 20 daughters with phenotypic records.
• These advancements hold significant promise for accelerating genetic gain in livestock populations.