Genetic gain & Inbreeding

Theory vs Simulation vs Reality: Predicting the outcomes of breeding programs

---

Abstract

Managing and designing a breeding program is a complex task that requires monitoring and predicting the outcomes of breeding to obtain high short-term genetic gain while maintaining long-term genetic diversity. Breeding goals differ across breeding programs but all rely on shared fundamental genetic principles that enable the development of predictive tools. On a population scale, quantitative genetic theory provides deterministic formulas to predict key metrics of breeding programs, such as the response to selection through the Breeder's equation. With rising computational power, stochastic simulations have become a powerful alternative to such inference by generating a digital twin that explicitly models phenotyping, mating, and breeding value estimation. While deterministic formulas are a powerful tool to easily approximate outcomes of a breeding program, stochastic simulations allow a much more in-depth analysis to isolate the impact of specific changes in breeding design to gain insights into how and where genetic gain is achieved or what variance in outcomes to expect. Simulation studies pose several practical challenges, including high computational demands that require simplifying reality, as a fully accurate simulation of, say, five years of breeding would require performing all computational operations, including all breeding value estimations that would be executed in that time frame. In this work, we discuss how to set up a simulation study, common issues and misconceptions with stochastic simulations, and how to avoid the latter, illustrated through the simulations of commercial breeding programs by the partners of the Breed4Food consortium. Exemplary consider here a commercial pig breeding program, aiming to analyze the impact of additional phenotyping and genotyping of crossbred animals in a test environment. This test environment is used as it more closely resembles the market environment than the nucleus environment, where purebred animals are housed. Both stochastic and deterministic prediction approaches suggest increased genetic gains of 10-12% when generating additional phenotypes and 30-35% when generating both phenotypes and genotypes, while real-world validation from previous studies suggests limited benefits of genotyping crossbred animals. Potential reasons for these biases are examined and discussed, e.g., with the integration of non-additive trait architectures reducing the added value of genotyping, and how much genetic gain from the nucleus translates to the market. Stochastic simulations furthermore provide a powerful tool to validate and complement analysis based on real-world experiments, as in the non-additive setting, lower genetic gain in purebreds does not necessarily imply lower genetic gains in crossbreds, highlighting the importance of distinguishing between the purebred and crossbred breeding value of an animal.

Keywords: 2026

How to Cite:

Pook, T., Jamieson, D., Perez, B., Orrett, C., Gil, M. & Hanenberg, E., (2026) “Theory vs Simulation vs Reality: Predicting the outcomes of breeding programs”, World Congress on Genetics Applied to Livestock Production Digital Archive 2026(1): 2252485. doi: https://doi.org/10.31274/wcgalp.23385

Rights: 1