Estimation & Prediction

Impact of temporal and kinship-based data truncation strategies on variance component estimates in genomics evaluation

---

Abstract

In routine genetic evaluations, strategically reducing data volume can lower computational costs without substantially compromising genetic variability or inference. In this context, we evaluated the impact of two truncation strategies on variance‐component estimation on real breeding programing for economically relevant productive and reproductive traits in a Nellore population under ~40 years of selection. The traits analyzed were 18-month weight (W18), scrotal circumference (SC), and probability of pregnancy at 14 months (PP14). The initial dataset comprised 578,186 phenotyped animals, 18,200 genotyped animals, and 653,095 pedigree records. Phenotypes spanned 40 years (1984-2024) from a predominantly closed Agro-Pecuária CFM population. We tested (i) temporal truncation using rolling 5-year windows (1984-2021) and (ii) sire truncation based on average kinship (0%, 10%, 30%) with the removal of sires and their offspring under three targeting rules: TOP (highest kinship), EXTREME (half highest + half lowest), and BOTTOM (lowest). Single-trait mixed models were fitted for each scenario to re-estimate additive genetic variance (σu2) and heritability (h2), including fixed contemporary group defined as herd, birth year/season, sex, and management group, when not modeled as random uncorrelated effect and random effects comprised additive genetic, and residual. The PP14 was analyzed with a Bayesian threshold model, whereas W18 and SC were fitted with linear animal models via AI-REML, all analyses were conducted using the BLUPF90 suites. Under rolling 5-year windows, σu2 remained temporally stable across traits, with an average stepwise variation between consecutive windows of 2.52% (SC), 2.18% (W18), and 8.41% (PP14). This supports the preservation of h2 and indicates that temporal data pruning can reduce computational demands with minimal impact on parameter estimation. In contrast, kinship-based sire truncation (0%, 10%, 30%) reduced σu2 for all traits, with larger impacts for W18 and SC (≈15-19% at 30%) and a marked reduction for PP14 when truncating TOP (−21%). These patterns may suggest that it is not sample size per se but the pedigree structure and sire representativeness in this, predominatly closed, Nellore population that sustains additive genetic variability. Focused pruning of highly related families compresses the genetic scale even when h2 shifts little, potentially diminishing response to selection and the robustness of future evaluations. We, therefore, recommend prioritizing temporal pruning and managing kinship representativeness to design more computationally efficient evaluations without loss of essential information. Moreover, truncation rules should be applied cautiously to avoid eroding groups carrying disproportionate shares of σu2, particularly for reproductive traits in Nellore breeding programs.

Keywords: 2026

How to Cite:

de Carvalho, F., Rocha, A., Polizel, G., Bis, F., Gubiani, G., Leão, M., Almeida, C., Campos, G., Ferraz, J., Brito, L., de Oliveira, E., Baldi, F. & Santana, M., (2026) “Impact of temporal and kinship-based data truncation strategies on variance component estimates in genomics evaluation”, World Congress on Genetics Applied to Livestock Production Digital Archive 2026(1): 2286722. doi: https://doi.org/10.31274/wcgalp.24105

Rights: 1