Genetic gain & Inbreeding

Assessment of reliabilities of estimated breeding values in pigs

---

Abstract

Reliability of estimated breeding values (EBVs) is a key parameter in genetic selection programmes, especially when it varies across candidates and traits. The main objective of this study is to compare results from the exact method using pedigree and genotype data with two approximate methods-pedigree only (PED) and pedigree plus genotype (ssSNP). Animals were differentiated into cohorts based on the availability of own phenotypic performance, genotype, and presence of progeny. The analysis used a dataset from a commercial Pietrain boar line within the Hendrix Genetics Swine breeding programme, including pedigree, genotype, and phenotype information for multiple traits with varying heritability and record numbers. Exact reliability was calculated from the diagonal elements of the inverse of the mixed model coefficient matrix, implemented in MiXBLUP. Due to matrix inversion limits, the exact method was capped at 40K genotyped animals. The dataset comprised 153K pedigree records, with 40K genotyped individuals. Three traits were studied: daily gain (DG), gestation length (GL), and feed conversion index (FCI), with 91.1K, 8.8K, and 16.6K records respectively. Heritability estimates were 0.40 (DG), 0.31 (GL), and 0.07 (FCI). The PED approximation followed Tier and Meyer's algorithm, ignoring genotype data, while the ssSNP method applied Gao et al.'s (2023) algorithm to incorporate genotypes. Both approximations were compared against exact reliabilities from single-trait models. As expected, results showed that genotype data markedly improved EBV reliability, particularly for traits with low heritability or in the absence of own or progeny performance. Without phenotypic records, adding genotypes increased reliability from 0.41 to 0.78 (DG), 0.40 to 0.67 (GL), and 0.15 to 0.46 (FCI). With phenotypic data, gains were smaller but remained notable for FCI. The ssSNP method aligned closely with the exact method for genotyped animals, showing high correlation and minimal bias, except for mild inflation bias in FCI. For ungenotyped animals, correlations were lower and slight regression bias appeared for DG in cohorts without progeny. The PED method consistently underperformed compared to ssSNP, even for ungenotyped animals, illustrating the limitations of pedigree-only approximations. The ssSNP method overcomes limitations of the exact method, allowing large scale application which would provide practical guidance for optimising resource allocation between phenotyping and genotyping to enhance selection accuracy in pig and other livestock breeding programmes.

Keywords: 2026

How to Cite:

Bink, M., ten Napel, J., Vandenplas, J. & Huisman, A., (2026) “Assessment of reliabilities of estimated breeding values in pigs”, World Congress on Genetics Applied to Livestock Production Digital Archive 2026(1): 2283155. doi: https://doi.org/10.31274/wcgalp.23490

Rights: 1