Study of Genetic Progress in the Context of Disconnection Between Two Originally Connected Populations.

Genetic progress of breeding programs is highly dependent on the size of reference populations and the relatedness between reference populations and selection candidates. Many reasons can lead a population to split into several subpopulations (sanitary, physiological, political reasons, etc.). More specifically, alternative (e.g., organic) farming may lead to farms breaking away from the conventional scheme to form a distinct breeding scheme, especially in organic sheep farming where the ban on hormones makes the use of artificial insemination (AI) difficult. However, these potential splits of the population into several smaller subpopulations could decrease genetic progress. The aim of our study was to investigate, using stochastic simulations, the impact of separation of the population into two subpopulations while still applying the same breeding objective and methods. We simulated a breeding program inspired by a dairy program but applicable to different species. We simulated two different initial population sizes with 5400 (10,800) females mated to 90 (180) males and a trait of heritability 0.30. This population was under selection for several discrete generations (G-9 to G-1) as a single population. Then, for the last 11 cycles of selection, the population was either maintained as a unique population (scenario "NoSep", which was the reference scenario) or split into two subpopulations with different ratios: 50/50, 60/40, 70/30, 80/20, and 90/10. We studied three scenarios in which the population was split: CE (separation and Common Evaluation), in which the evaluation remained common between both subpopulations; SE (separation and separate evaluation), in which the subpopulations were evaluated individually; and NoSel (Separation and No Selection), in which the breeding males were randomly selected, as opposed to the two previous scenarios in which we selected the males based on their GEBVs. We studied the evolution of differentiation of populations (F_st), accuracy of predictions, genetic progress, and rate of inbreeding over generations. We observed a faster genetic divergence in the case of an unbalanced split and separate evaluation (F_st in G11 equal to 0.134 for the ratio 90/10 scenario SE). The separate evaluation had a significant, negative effect on both the accuracy and genetic gain of the smallest population (minimal accuracy of 0.53 and maximal loss of 16.6% for ratio 90/10 with 5400 females), whereas the accuracy and genetic gain of the largest population were not impacted. Combining the evaluations led to smaller but still significant deterioration of the genetic gain of the smallest population when the ratio was very unbalanced (loss of genetic gain of 14.3% for a ratio of 90/10 with 5400 females). In conclusion, population separation has a negative impact on genetic gain, particularly for small populations. Although it does help in alleviating divergence and loss of genetic gain, joint evaluation can not fully compensate for the split of the populations.