Global breeding programs

Impact of data availability and genetic connectedness on heritability estimates: a case study

---

Abstract

Accurate and reliable estimation of variance components is crucial in animal breeding programmes, as it guides selection and conservation strategies. Estimation can be particularly complicated in small-scale breeding programmes at the start of genomic selection or during the transition from pedigree-based to genomic selection. Specifically, many livestock populations, especially indigenous breeds, often lack the structure and data quality needed for robust estimation. For example, they struggle with small flocks/herds with low connectedness, unbalanced availability of phenotypes and genotypes, and inaccurate or even completely non-existent pedigrees. Our objective was to explore how data availability and genetic relationships affect variance components estimation and the reliability of estimated breeding values across farms of varying sizes. In this case study, we evaluated heritability estimates for milk, fat and protein yields in a local Croatian sheep breed using three distinct approaches: pedigree-based residual maximum likelihood (REML), genomic-based single-step REML (ssREML), and Bayesian inference via integrated nested Laplace approximation (INLA). Furthermore, we aimed to examine whether accounting for spatial relationships between neighbouring flocks impacts estimates. Our dataset was comprised of about 33,000 whole-lactation records from 10,308 Pag sheep ewes spread across 107 flocks. After quality control, genomic information was available for 2,134 sheep (2,031 females and 103 males) and for 46,912 single nucleotide polymorphism markers. Out of 2,031 genotyped females, 1,792 ewes from 46 flocks were amongst those with whole-lactation records. Pedigree was trimmed to include all phenotyped and genotyped individuals, along with three generations of their ancestors without phenotypic or genotypic information, yielding a total of 12,257 animals. Pedigree was processed with renumf90 software, and REML and ssREML analyses were performed with blupf90+ software. Bayesian inference was performed using the R-INLA package, which allows spatial modelling of random, correlated flock effects using the Matern covariance function based on Euclidean distances between flocks. Connectedness analysis was based on the coefficient of determination and was performed using the GCA R-package. Heritability estimates obtained after removing phenotypes from low-, mid-, or highly connected animals were compared with full-data estimates, focusing on standard errors and changes in individual animal reliabilities and estimated breeding values (EBV) as stability metrics. Our results demonstrate that while pedigree-based estimation is reliable for well-connected flocks with many individuals, it becomes unreliable for smaller and more fragmented flocks. In such cases, genomic-based estimation, which captures additional genetic relationships, provides more reliable heritability estimates. Furthermore, connectedness between flocks raises important considerations about how the breeding scheme should be managed and adjusted, as heritability estimates and the EBV reliability were more sensitive in low-connected flocks than in well-connected flocks. Finally, this study contributes to optimising cost-effective phenotyping and genotyping strategies when large-scale actions are not feasible due to economic and logistical constraints.

Keywords: 2026

How to Cite:

Pocrnič, I., Rochus, C., Ramljak, J., Ceranac, D., Barač, Z., Špehar, M. & Kasap, A., (2026) “Impact of data availability and genetic connectedness on heritability estimates: a case study”, World Congress on Genetics Applied to Livestock Production Digital Archive 2026(1): 2275068. doi: https://doi.org/10.31274/wcgalp.23422

Rights: 1