Genetic gain & Inbreeding

Comparing genetic trends across models and implementations

---

Abstract

The introduction of single step GBLUP or modifications in models or software may lead to changes in Estimated Breeding Values (EBVs). For instance, data editing, redefining the contemporary groups, use of Unknown Parent Groups (UPG) or Metafounders (MF), or changes in variance components will result in different genetic trends. Large changes are not expected (original models are usually reasonably good) but small ones are expected (refined models should give more accurate (or less biased) evaluations). The objective is to show a manner of exploring these differences.For traits with several years of recording, the analyst wishes to compare genetic trends. However, comparing shapes is difficult - for instance, how much different is a "U" from a "V" or a "~" from a "-" ? Because visual comparison of trend shapes is subjective, here we propose not to compare the genetic trends themselves, but to compare the annual genetic gains - the DeltaGs. True DeltaG will depend on factors such as selection intensity and accuracy of evaluation. Estimated DeltaG will depend foremost on the ability of the model to split a phenotypic trend into environmental and genetic trends. Visual inspection of DeltaGs is easier because successive years are made independent from each other.A simple way of comparing genetic trends is therefore to split them into yearly DeltaG's, and then compare those, for instance, via correlation. We believe that comparing DeltaGs gives good initial indications which are useful for exploration and debugging.We illustrate this with two traits from US domestic pedigree-BLUP evaluations run by the Council on Dairy Cattle Breeding (CDCB). We considered two different models. The first model (UPG) fits UPGs conventionally. The second model (MF) uses MF, with a proper prior relationship matrix among them, and compacts MF definitions until they gather enough information (Legarra and Aguilar, 2024). In Holstein, the correlation between DeltaGs for milk yield among proven bulls was 1 with an almost overlapping genetic trend. However, the same two models fit for heifer livability resulted in a correlation across DeltaGs of 0.96. Whereas both models showed positive trends, MF produced slightly but consistently smaller yearly DeltaGs, yielding a lower overall genetic trend. These particular differences of MF vs UPG differences for heifer livability warrant further investigation.To conclude, correlating yearly DeltaGs provides an extra tool for exploring and debugging evaluation changes.

Keywords: 2026

How to Cite:

Legarra, A., Nicolazzi, E. & Savoia, S., (2026) “Comparing genetic trends across models and implementations”, World Congress on Genetics Applied to Livestock Production Digital Archive 2026(1): 2285373. doi: https://doi.org/10.31274/wcgalp.23663

Rights: 1