Estimation & Prediction

Temporal stability of indirect predictions in single-step genomic models

---

Abstract

Genomic breeding values (GEBV) from genetic evaluations using single-step GBLUP with the algorithm for proven and young (APY-ssGBLUP) can be back-solved to obtain SNP effects. These are useful for computing indirect predictions (IP) in genotyped animals not included in official evaluations (i.e., young selection candidates, or commercial animals). As new data are added to the evaluations, SNP effects need to be updated to reflect changes in the population's genetic makeup. This study aimed to 1) compare two algorithms (A1 and A2) for obtaining IP from APY-ssGBLUP models and 2) assess changes in IP over two years in a large beef cattle population. While both algorithms use APY, A1 calculates marker effects based on GEBV from all genotyped animals, whereas A2 is based on GEBV from core animals. The American Angus Association provided three datasets: growth, docility, and feed efficiency. The pedigree file contained 12 million animals, of which 1.8 million were genotyped. A total of 40,000 young, genotyped animals, born between 2018 and 2022, were chosen as the focal group. GEBV from APY-ssGBLUP models, including data up to December 2022, served as a benchmark for evaluating the similarity between IP and GEBV in the focal group. Then, IP were obtained using the same dataset, but with the genotypes and pedigree records from the focal group removed. To address the second objective, IP for all evaluated traits were calculated across three time points: December 2022, December 2023, and December 2024. Mean (max) absolute difference between IP and GEVB ranged from 0.11 (0.63) to 0.91 (2.58) additive genetic SD (SDa) across all traits using A1, whereas for A2 it ranged from 0.07 (0.36) to 0.70 (1.28) SDa. Correlations between IP obtained across the three years ranged from 0.95 to 0.99 in A1 and from 0.96 to 1.00 in A2. The lowest correlations were observed for traits with smaller reference populations (i.e., docility and feed efficiency). Regardless of the method, the correlation between IP over a one-year period was consistently above 0.97. However, A2 provided more stable IP over the period of two years. Overall, IP calculated based on SNP effects backsolved from GEBV for core animals in APY-ssGBLUP remains stable over time in large populations.

Keywords: 2026

How to Cite:

Sanchez-Sierra, S., Garcia, A., Medeiros, R., Retallick, K., Bermann, M. & Lourenco, D., (2026) “Temporal stability of indirect predictions in single-step genomic models”, World Congress on Genetics Applied to Livestock Production Digital Archive 2026(1): 2280557. doi: https://doi.org/10.31274/wcgalp.23445

Rights: 1