Breeding goals & design

A practical tool: automated pipelineà¢Ë†â€™genetic evaluation quality control and auto-fitting into non-linear selection index

---

Abstract

Once a new selection index has been integrated into a genetic evaluation (GE) pipeline, it usually is applied to subsequent evaluation runs without detailed checking. It is becoming increasingly common to integrate non-linear profit functions (NLPF) in index formulations which commonly rely on thresholds at specific positions on the estimated breeding value (EBV) scale. Therefore, when the EBV distribution or the genetic base changes, it is highly likely that the NLPF will need to be adjusted or recalculated. The objective here is to describe an automated pipeline to discover any significant deviation of new EBV from previous runs and standardize the new EBV to the previous scale so that the developed NLPF can be applied to the new EBV without manual adjustments. Two datasets were used in this study. Firstly, New Zealand national dairy evaluations in 2023 and 2025 with a quadratic NLPF created in 2023 for udder overall where the profit ($) plateaued when the EBV reached 1.15 (score). Secondly, New Zealand high merit beef evaluations in 2019 and 2024 with a piecewise NLPF created in 2019 for gestation length where the profit was $0 when the EBV was smaller than -4 days but increased linearly afterwards. Methods for quality control (QC) included comparison of means, variances and linear regressions of previous and new EBV. The method first standardized the new EBV by regressing them towards the previous EBV, then applied the previously developed NLPF to the standardized EBV. The QC pipeline detected a base adjustment of the mean in 2025 EBV compared to 2023 in the common animals across evaluations of dairy data (0.05±0.45 vs. 0.17±0.45, Pgref=0.988gnew+0.125 (adj-R2=0.99) where gref and gnew were EBV of common animals in the old and new evaluations, respectively, indicating little deviation of EBV across evaluations. Using this regression to standardize gnew and fitting into the old NLPF, we obtained a new quadratic NLFP for new EBV with different coefficients and EBV threshold (1.03 score). For high merit beef data, although there was no significant difference between the new and old EBV (0.14±1.41 vs. 0.17±1.58, P=0.77), the regression of the old on new EBV in the common animals was gref= 0.655gnew+0.027 (adj-R2=0.54). The low slope and adj-R2 values indicated a deviation across GE. Fitting this into the old NLPF, we obtained a new piecewise NLPF with different coefficients and a threshold at -6.15 days. In conclusion, the proposed pipeline automated the QC of new GE and the generation of new non-linear index values. Once adopted, this can increase quality check frequency, prevent errors by automatically applying NLPF to standardized EBV, and reduce manual workload associated with recalibrating non-linear indexes.

Keywords: 2026

How to Cite:

Zhang, X., Hely, F., Stachowicsz, K. & Amer, P., (2026) “A practical tool: automated pipelineà¢Ë†â€™genetic evaluation quality control and auto-fitting into non-linear selection index”, World Congress on Genetics Applied to Livestock Production Digital Archive 2026(1): 2282976. doi: https://doi.org/10.31274/wcgalp.23473

Rights: 1