360. Indirect genomic prediction reduces computational costs in large-scale single-step evaluations

Abstract

Computing time and memory requirements increase with single-step methods to estimate genomic breeding values when the number of genotyped animals increases. Computational costs can be reduced by omitting genotypes of animals without phenotype and progeny, often the candidate animals for selection. Indirect prediction of a candidate animal GEBV can be based on the animal's genotype and SNP marker solutions (DGV). Alternatively, the sum of DGV and residual polygenic (RPG) effect can be computed, denoted GRV (Genomic and Residual polygenic Value). We applied indirect genomic prediction for a 6 trait calving difficulty evaluation. There were 1.50 million genotyped animals of which 36% were considered candidate animals. Based on our results, DGV showed high accuracy but also bias due to omitting the RPG effect. The GRV prediction had high accuracy and low bias. Computing time was reduced by 33%.

Derivation of indirect predictions using genomic recursions across generations in a broiler population

“…An important point to highlight is that in our study, to derive the IP we ignored the residual polygenic effect contribution, which has a marginal effect as the proportion assumed was 5% (i.e., ); however, when the residual polygenic effect fitted in the statistical model is larger, its contribution is more significant. In such a case, it is important to consider a more exact IP derivation by Liu et al (2016) and recently investigated by Strandén et al (2022) and Vandenplas et al (2023) .…”

“… Strandén et al (2022) used calving difficulty data from the Irish Cattle Breeding Federation. For the trait with the largest benefit, accounting for the residual polygenic effect ( =0.20 in their study) increased the correlation between GEBV and IP from 0.966 to 0.996.…”