Genetic gain & Inbreeding

Comparison of genetic gain predicted in swine population using stochastic and determinist simulation with real gain observed

---

Abstract

Breeding programs are becoming increasingly complex due to the integration of multi-trait objectives and multi tools approaches. Moreover, current breeding goals now extend beyond direct economic values to also incorporate traits related to welfare and adaptation to climate change. To predict genetic trends under different breeding goals, two simulation approaches are available: deterministic or stochastic. The study aimed to compare genetic trends predicted by both simulation approaches with the real genetic evolution observed in the Gamma pig population, a synthetic maternal line. Genetic trends were analyzed for five traits: number of born alive piglets (BA), average piglet birth weight (ABW), age at 100 kg (A100), backfat (B100) and loin depth (L100) estimated at 100 kg. Simulations were performed using two software packages: Selaction (deterministic, D), and AlphaSimR (stochastic, S), with identical trait sets, source of information, genetic parameters, population structure and selection intensity. The real genetic evolution was estimated from genetic merit of animals born between 2023 and 2025. Results were expressed in genetic standard deviation for each trait and compared to the real genetic gain observed as percentage of change. For production traits (B100, L100), the D method closely matched the observed gain, with less than 1% difference for B100 and 12% for L100. In contrast, the S method overestimated the gains by 45% and 25%, respectively. For BA, the S method predicted a genetic improvement close to observes trend, 0.9% difference, while the D method underestimated the gain by 46%. Both methods overestimated the gentic gain for ABW and A100. These results suggest that the deterministic approach is best suited for traits with large number of phenotyped canidates and with strong selection intensity, particularly production traits, whereas the stochastic method tended to overestimate gains. Conversely, the stochastic method was more accurate for complex traits such as BA, wich are phenotyped after the main selection step at 100 kg, and measured in only one gender, where the D method underestimated the gain. The similar results obtained for ABW and A100, despite their differing suitability for each method, may be explained by strong genetic correlations, favourable between ABW and A100, two traits related to growth, and unfavourable with BA. In conclusion, both simulation methods are valuable tools to predict genetic gains. Deterministic methods offer reliable results for high intensity selection steps and stochastic methods providing greater accuracy for complex traits such as repeated measurement after main selection step. Given their computational differences, these two approaches can be used complementarily to evaluate the impact of selection strategies.

Keywords: 2026

How to Cite:

Cudrey, D., Flatres-Grall, L., Kashefifard, K. & Lenoir, G., (2026) “Comparison of genetic gain predicted in swine population using stochastic and determinist simulation with real gain observed”, World Congress on Genetics Applied to Livestock Production Digital Archive 2026(1): 2284538. doi: https://doi.org/10.31274/wcgalp.23581

Rights: 1