The effect of nucleus size and selection intensity on the amount of bias due to ignoring the effect of selection in estimated genetic gain in nucleus breeding schemes in Lori Bakhtiari sheep

One of the breeding programs for sheep is the nucleus breeding scheme (Kosgey and Okeyo 2007). The nucleus breeding system, including open or close nucleus, is a good start for genetic improvement in many populations. Unlike the close nucleus schemes, there is a reciprocal gene flow in open nucleus schemes, where the best animals from base layer could be transferred to the nucleus layer (Kosgey et al. 2006). In the original nucleus breeding scheme, a constant genetic variation was assumed for the trait of interest across different generations (Hopkins and James 1978). However, it is well realized that the additive genetic variance is decreased after several generations of intense selection. Thus, the actual genetic gain is likely less than the expected genetic gain. In other words, the outcomes of these breeding schemes are likely affected by the intensity of selection as well as the relative size of the nucleus population to the size of whole population. The aim of current study was to investigate the effect of relative nucleus size at three levels of 5, 10 and 15 percent and the effect of selection intensity in males and females at three levels of high (i.e. selection proportion in male and female candidates were 1 and 20 percent, respectively), moderate (i.e. selection proportion in male and female candidates were 5 and 50 percent, respectively) and weak (i.e. selection proportion in male and female candidates were 10 and 80 percent, respectively) on the amount of biasness of predicted genetic gain due to ignoring the effect of selection on genetic variance. An open nucleus breeding scheme was deterministically simulated using data from Lori Bakhtiari sheep breed, consisting a nucleus herd of 1005 ewes and a base herd of 4019 ewes (Vatankhah 2005). Birth weight was the breeding goal trait and selection was carried out using one individual performance record. The generation was assumed to be overlapped and three and six age groups were considered for males and females, respectively. The age of parents at first lambing was 2 years and 0.94 yearling lamb was produced per ewe. One ram was mated to 33 ewes. The additive genetic variance for breeding goal trait was 0.11 and its heritability was set to 0.30. For a given size of the nucleus and a constant rate of transferring ewes from base population to nucleus (x), there was an optimum rate of migration of nucleus rams into the base population (w), that maximized the expected genetic gain and minimized the genetic lag between nucleus and base populations, with or without accounting for the effect of selection on the genetic variance. By increasing the relative size of the nucleus, from 5 to 15 percent of whole population, the biasness of the predicted genetic gain was increased from 11.84 to 16.67 percent. At a given rate of x, the bias of genetic gain prediction which originated from ignoring the effect of selection, varied nonlinearly as the rate of w increased and had a minimum value at the optimal value of w. At a particular rate of w, increasing the rate of x, was associated with higher rate of bias. The rate of bias increased from 12.2 to 16.54 percent, as the intensity of selection increased. The reason could be the markedly reduction in the genetic variance of considered trait due to selection at high selection intensities. In open nucleus schemes, the amount of bias changed from 14.18 to 17.72, as a consequence of increasing the intensity of selection. Across all levels of selection intensity in open nucleus programs, the lowest bias was observed at the rate of 0.75 for w. Evaluation of joint effect of relative nucleus size and intensity of selection, showed that at high selection intensity and relative nucleus sizes higher than 10 percent, the amount of bias was not likely to be influenced by nucleus size. Ignoring the effect of selection on genetic variance could result in a biased prediction of genetic gain in the nucleus breeding schemes. However, the optimal ram transfer rate from nucleus to base population was independent from the effect of selection intensity. The lowest bias was observed for the scheme with optimum transfer rate. Results of the current study revealed that at weak selection intensity, the effect of selection could be ignored and the breeding program could be optimized for ram transfer rate from nucleus to the base populations. However, at high selection intensities, the ignorance of this effect might result in wrong conclusions.