Genomic selection for multiple maternal and growth traits in large white pigs using Single-Step GBLUP

• Published in: Journal of animal science Vol. 97; p. 42
• Main Authors: Fragomeni, Breno, Vitezica, Zulma, Liu, Justine, Huang, Yijian, Gray, Kent, Lourenco, Daniela, Misztal, Ignacy
• Format: Journal Article
• Language: English
• Published: Champaign Oxford University Press 01.12.2019
• Subjects: Algorithms; Animal breeding; Animal models; Animal populations; Breeding of animals; Computational efficiency; Computing time; Correlation; Genetics; Genomics; Genotype & phenotype; Hogs; Livestock; Mortality; Phenotypes; Ranking; Single-nucleotide polymorphism; Swine; Weaning; Weight; United States > US; Georgia
• ISSN: 0021-8812; 1525-3163

Genomic selection for multiple maternal and growth traits in large white pigs using Single-Step GBLUP. Breno Fragomeni1, Zulma Vitezica2, Justine Liu3, Yijian Huang4, Kent Gray5, Daniela Lourenco6, Ignacy Misztal7, 1, 2, 3University of Connecticut, 4Smithfield Premium Genetics, 5Smithfield Premium Genetics, 6University of Georgia, 7University of Georgia The objective of this study was to implement a multi-trait genomic evaluation for maternal and growth traits in a swine population. Phenotypes for preweaning mortality, litter size, weaning weight, and average daily gain were available for 282K Large White pigs. The pedigree included 314k individuals, of which 35,731 were genotyped for 45K SNPs. Variance components were estimated in a multi-trait animal model without genomic information by AIREMLF90. Genomic breeding values were estimated using the genomic information by single-step GBLUP. The algorithm for proven and young (APY) was used to reduce computing time. Genetic correlation between proportion and the total number of preweaning deaths was 0.95. A strong, positive genetic correlation was also observed between weaning weight and average daily gain (r = 0.94). Conversely, the genetic correlations between mortality and growth traits were negative, with an average of -0.7. To avoid computations by expensive threshold models, preweaning mortality was transformed from a binary trait to two linear dam traits: proportion and a total number of piglets dead before weaning. Because of the high genetic correlations within groups of traits, inclusion of only one growth and one mortality trait in the model decreases computing time and allows for the inclusion of other traits. Reduction in computing time for the evaluation using APY was up to 20x, and no differences in EPD ranking were observed. The algorithm for proven and young improves the efficiency of genomic evaluation in swine without harming the quality of predictions. For this population, a binary trait of mortality can be replaced by a linear trait of the dam, resulting in a similar ranking for the selection candidates.