32 A Bayesian Hierarchical Model to Integrate Growth Models into Genomic Evaluation of Pigs. (8th October 2021)

• Record Type: Journal Article
• Authors: Yu, Haipeng; Milgen, Jaap; Knol, Egbert; Fernando, Rohan; Dekkers, Jack C
• Is Part Of: Journal of animal science. Volume 99(2021)Supplement 3
• Page Start: 18
• Page End: 19
• Publication Date: 2021-10-08
• Subjects: Bayesian hierarchical model -- pig growth model -- genomic prediction; Livestock -- Periodicals; Livestock; Electronic journals; Periodicals; 636.005
• DOI: 10.1093/jas/skab235.030
• Languages: English
• ISSNs: 0021-8812

Abstract

Abstract: Genomic prediction has advanced genetic improvement by enabling more accurate estimates of breeding values at an early age. Although genomic prediction is efficient in predicting traits dominated by additive genetic effects within common settings, prediction in the presence of non-additive genetic effects and genotype by environmental interactions (GxE) remains a challenge. Previous studies have attempted to address these challenges by statistical modeling, while the augmentation of statistical models with biological information has received relatively little attention. A pig growth model assumes growth performance is a nonlinear functional interaction between the animal's genetic potential for underlying latent growth traits and environmental factors and has the potential to capture GxE and non-additive genetic effects. The objective of this study was to integrate a nonlinear stable Gompertz function of three latent growth traits and age into genomic prediction models using Bayesian hierarchical modeling. The three latent growth traits were modeled as a linear combination of systematic environmental, marker, and residual effects. The model was applied to daily body weight data from ~83 to ~186 days of age on 4, 039 purebred boars that were genotyped for 24K markers. Bias and prediction accuracy of genomic predictions of selection candidates were assessed by extending the linear regression method of predictions based on part and whole data to a non-linear setting. The accuracy (bias) of genomic predictions was 0.58 (0.82), 0.46 (0.90), 0.54 (0.78), and 0.60 (0.84) for the three latent growth traits and average daily gain derived from integrated nonlinear model, respectively, compared to 0.58 (0.87) for genomic predictions of average daily gain using standard linear models. In subsequent work, the growth model will be extended to include daily feed intake and carcass composition data. Resulting models are expected to substantially advance genetic improvement in pigs across environments. Funded by USDA-NIFA grant # 2020-67015-31031.