The Impact of Truncating Data on the Predictive Ability of Selection Candidate EBV in Swine Using Ssgblup.

As generational information accrues in a population, selection candidates become more distantly related to the majority of the historic population, thus eroding the benefit of data from historic animals relative to predicting the genetic merit of selection candidates. Therefore, the objective was to assess the impact of removing older data on the predictive ability of selection candidate EBV in Duroc and Yorkshire swine populations. Within each population backfat (BF), loin eye depth (LED) and average daily gain in the nursery (ADGn) were investigated. For Yorkshire, the data consisted of phenotyped animals (n = 95,769-117,875) born from 2005 to 2017 along with 5,783 genotyped animals. For Duroc, phenotyped animals (n = 135,324-152,302) born from 2003 to 2017 along with 12,180 genotyped animals were utilized. Genotyped animals were born from 2011 to 2017. Within trait, a ssGBLUP model was utilized and the pedigree was truncated to 3 ancestral generations back from phenotyped animals. The impact of removing older data was determined by masking the phenotype when predicting the estimated breeding value (EBV) for animals born recently (i.e. 2016-2017). The phenotypes and genotypes were iteratively removed one year at a time starting with the oldest year and ending with 2015. Corrected phenotypes (Cp) were estimated by adjusting for fixed and random environmental effects. The change in the EBV predictive ability when removing data was determined as the correlation between Cp and EBV. Removing phenotypes from animals born prior to 2011, 34 and 32% of the data for Duroc and Yorkshire, respectively, resulted in a negligible or a slight numerical increase in the correlation between Cp and EBV. For Duroc, the correlations ranged from 0.0750 to 0.1771 when utilizing all data and on average increased by 0.0017 when removing data prior to 2013 when the largest correlations were estimated. For Yorkshire, the correlations ranged from 0.0743 to 0.1985 when utilizing all data and on average increased by 0.001 when removing data prior to 2011 (BF) or 2012 (ADGn & LED) when the largest correlations where estimated for these traits. Lastly, as older animals were removed, the correlation between pedigree and genomic based relationships for animals with genotypes increased. The current study has demonstrated that removing phenotypic information on older animals resulted in a numerically higher EBV predictive ability for selection candidates.