Genomic prediction of disease occurrence using producer-recorded health data: a comparison of methods

• Published in: Genetics selection evolution (Paris) Vol. 47; no. 1; p. 41
• Main Authors: Parker Gaddis, Kristen L, Tiezzi, Francesco, Cole, John B, Clay, John S, Maltecca, Christian
• Format: Journal Article
• Language: English
• Published: France Springer-Verlag 08.05.2015; BioMed Central Ltd; BioMed Central
• Subjects: Analysis; Animal breeding; animal health; Animal husbandry; Animal lactation; Animals; Bayes Theorem; Bivariate analysis; bovine mastitis; Breeding - methods; breeding value; Cattle; Dairy cattle; Dairy industry; Dairying; data collection; Datasets; Daughters; disease occurrence; Female; Fertility; Fitness; genetic correlation; Genetic factors; Genetic research; Genomics; Genomics - methods; Genotype & phenotype; Heritability; Liability; Life Sciences; marker-assisted selection; Mastitis; Mastitis, Bovine - epidemiology; Mastitis, Bovine - genetics; Methods; Performance evaluation; Population; prediction; Predictions; Reliability aspects; Reproductive fitness; Selection, Genetic; sires; somatic cells; variance; United States > US
• ISSN: 1297-9686; 0999-193X; 1297-9686
• DOI: 10.1186/s12711-015-0093-9

BACKGROUND: Genetic selection has been successful in achieving increased production in dairy cattle; however, corresponding declines in fitness traits have been documented. Selection for fitness traits is more difficult, since they have low heritabilities and are influenced by various non-genetic factors. The objective of this paper was to investigate the predictive ability of two-stage and single-step genomic selection methods applied to health data collected from on-farm computer systems in the U.S. METHODS: Implementation of single-trait and two-trait sire models was investigated using BayesA and single-step methods for mastitis and somatic cell score. Variance components were estimated. The complete dataset was divided into training and validation sets to perform model comparison. Estimated sire breeding values were used to estimate the number of daughters expected to develop mastitis. Predictive ability of each model was assessed by the sum of χ²values that compared predicted and observed numbers of daughters with mastitis and the proportion of wrong predictions. RESULTS: According to the model applied, estimated heritabilities of liability to mastitis ranged from 0.05 (S D=0.02) to 0.11 (S D=0.03) and estimated heritabilities of somatic cell score ranged from 0.08 (S D=0.01) to 0.18 (S D=0.03). Posterior mean of genetic correlation between mastitis and somatic cell score was equal to 0.63 (S D=0.17). The single-step method had the best predictive ability. Conversely, the smallest number of wrong predictions was obtained with the univariate BayesA model. The best model fit was found for single-step and pedigree-based models. Bivariate single-step analysis had a better predictive ability than bivariate BayesA; however, the latter led to the smallest number of wrong predictions. CONCLUSIONS: Genomic data improved our ability to predict animal breeding values. Performance of genomic selection methods depends on a multitude of factors. Heritability of traits and reliability of genotyped individuals has a large impact on the performance of genomic evaluation methods. Given the current characteristics of producer-recorded health data, single-step methods have several advantages compared to two-step methods.