25 Determining stability of genomic predictivity in future generations in commercial pig populations

Abstract

Abstract Genomic predictivity is expected to decay over time as predictions are evaluated to more distant generations. More data increases predictive ability; however, data from distant ancestors may not add a significant amount of value compared to the data from closely related individuals. The objective of this study was to evaluate the decay in genomic predictivity over time and to compare the magnitude of decay when including ancestral data versus data from 2 and 3 most recent generations for body weight at off-test (BW). The data set included 211,812 phenotypic records. The pedigree included 406,983 animals from 2001 to 2020, of which 55,118 were genotyped. A single-trait model was used with all ancestral data and sliding subsets of two- and three-generation intervals. Single-step GBLUP was used to calculate GEBVs. Predictive abilities were calculated by the correlation between GEBVs and adjusted phenotypes. Validation populations consisted of single generations succeeding the training population and continued for all generations available. Predictive ability was slightly higher, with all ancestral data in the training population compared to three- and two-generation intervals. The decay of predictivity was similar when comparing the three training population subsets. The average predictivity for the validation population immediately following the training population was 0.40 for 2016, 0.39 for 2017, 0.35 for 2018, and 0.29 for 2019. Predictive ability reached a maximum in the year 2017 (0.45) for the ancestral training population, 2017 (0.43) for the 3-year training population, and 2016 (0.38) for the 2-year training population. The average decay of predictive ability from the first year after the training population to the second year was -0.08. Realized predictivity is affected by selection pressure. The drop in predictive ability suggests declining heritability. With more data and with consistent selection pressure, predictive abilities should increase.