Genetic gain & Inbreeding

Incorporating Prediction Error Variance-Covariance into Selection Processes to Enhance the Probability of Identifying Elite Products

---

Abstract

In genomic selection (GS), genomic breeding values (GEBVs) are estimated along with prediction error variances (PEV) and covariances (PEC) from a genomic best linear unbiased prediction (GBLUP) model. PEV quantifies uncertainty in GEBVs, while PEC reveals relationships among them. GS is widely used in plant breeding, focusing on two components: population improvement, which resembles nucleus population improvement in animal breeding, driven by the population mean through continuous cycling of progenies for subsequent crossings, and product development, which follows traditional plant breeding methodology that encompasses multiple selection stages. For product development, progenies are selected using GS after crossings to develop inbred lines, which are then subjected to years of pre-commercial testing to identify the best products. This study aims to determine the optimal number of individuals to select during GS to maximize the probability of advancing the best products. To maximize the probability of selecting the best products, both PEV and PEC from the GBLUP model, based on real soybean data, were considered alongside GEBVs. Ten thousand samples were drawn from a multivariate normal distribution, using GEBVs as the mean and PEV/PEC as the variance-covariance matrix for 1,000 genotyped individuals. This sampling aimed to estimate the probability of selecting at least 10 products that exceed a predefined yield threshold, based on truncation selection using GEBV rankings. Probabilities were calculated under various truncation selection intensities (ranging from 10 to 1,000), defined as the number of successes (at least 10 individuals exceeding the threshold) divided by the total number of samples from the selected individuals. The results show that as the number of selected individuals increases from 10 to 1,000, the probability of success rises from below 5% to a plateau of 20% at 600 selected individuals, and then remains around 20% even with the selection of all individuals. The optimal number of selected individuals is then identified as the point at which the probability levels off as the number of selected individuals increases. In conclusion, incorporating known PEV and PEC through simulations proved effective in determining the optimal number of individuals needed to maximize the probability of identifying elite products at a specified target threshold. These findings suggest that by integrating PEV and PEC into GS processes, breeders can enhance their selection strategies, ultimately leading to improved breeding outcomes.

Keywords: 2026

How to Cite:

Cheng, J., Gaynor, C., Lehermeier, C., Teyssedre, S., Garcia, D., Branston, K., Tseng, M., Hollifield, M., Iheshiulor, O. & Davis, S., (2026) “Incorporating Prediction Error Variance-Covariance into Selection Processes to Enhance the Probability of Identifying Elite Products”, World Congress on Genetics Applied to Livestock Production Digital Archive 2026(1): 2286035. doi: https://doi.org/10.31274/wcgalp.23841

Rights: 1