Proceedings of the 6th World Congress on Genetics Applied to Livestock Production
Vol. 25:455-458
Armidale, NSW, Australia

P. Uimari¹ and J. Gibson^2
1Rolf Nevalinna Institute, University of Helsinki
00014 University of Helsinki, Finland
²Centre for Genetic Improvement of Livestock
Department of Animal and Poultry Science
University of Guelph, Guelph, Ontario, N1G 2W1, Canada

SUMMARY

A stochastic simulation was used to investigate the value of crossbreeding information in a two line crossbreeding system in poultry. Populations consisted of 25 sires and 250 dams, which were mated to produce both purebred and crossbred progeny. The next generation parents were selected either based on purebred information (PLS) (sib means, and own performance for females) or additional crossbred sib means were included (CCPS). A trait under selection was controlled by 20 loci with varying degrees of dominance. Pure lines differed in initial allele frequencies. If the trait was controlled by loci with partial dominance no extra benefit was obtained from including crossbred information over pure line information. Under complete dominance and overdominance CCPS outperformed PLS. It was concluded that a breeder can maximize the rate of progress by collecting crossbred data but the benefits will likely be small if the trait under selection is controlled by genes with partial dominance.

INTRODUCTION

The commercial product in the poultry breeding is a crossbred bird. To maximize the performance of the final crossbred animals several breeding strategies are available including pure line selection (PLS), reciprocal selection (RS), reciprocal recurrent selection (RSS), or a combination of PLS and either RS or RSS (Hunton, 1990). Wei and van der Werf (1994) introduced a combined crossbred and purebred selection (CCPS). In CCPS the crossbred and purebred performances are combined using general selection index theory and a correlation term r_pc (correlation between crossbred and purebred performances) as a weighting factor. Wei and van der Werf (1994) found that CCPS was favored over both PLS and CP (direct selection of crossbred performance) in most of the cases studied. PLS was better than CCPS if the total number of purebred and crossbred progeny was fixed and r_pc was higher than 0.8.

Theoretical comparisons between CCPS and PLS presented by Wei and van der Werf (1994) were based on a theory of the infinitesimal model with additively inherited traits, which contradicts the assumption that the traits for which crossbred information is used in practice express significant non-additive variation. In this study, stochastic simulations, were used to compare CCPS and PLS, when inheritance is controlled by many loci with varying degree of dominance. In simulation studies several factors that cannot easily be taken into account in theoretical calculations, can be handled through stochastic processes including, changes in allele frequencies, genetic drift due to finite population size, and control of non-additive inheritance by many loci. An additional advantage of stochastic simulations is that the results can be expressed in terms of both mean and variance of response so that the risk involved in a particular selection scheme can be considered.

MATERIALS AND METHODS

A bi-allelic finite unlinked locus model (20 loci) allowing for dominance was used to generate trait observations. The degree of dominance was either 0.5, 0.75, 1.0, or 1.5. Genotypic values for each locus were the same for both lines, but frequencies of the favorable alleles were line dependent. For the first line the allele frequencies at successive loci were 0.7-0.3-0.7-0.3 … and for the second line the allele frequencies were 0.3-0.7-0.3-0.7 …, thus if the frequency of allele was high in line 1 it was low in line 2 to ensure heterosis. Genetic values for base population animals were generated under Hardy-Weinberg and gametic phase equilibrium and genetic values of offspring were generated using independent segregation of alleles between loci. Environmental factors were generated independently from a normal distribution with mean zero and variance equal to (1-H²)s _p², where H² is the broad sense heritability and s _p² is phenotypic variance. The trait simulated was assumed to be sex limited so only females had observations.

Populations consisted of 25 sires and 250 dams, which were mated to produce both purebred and crossbred progeny. Parents for the next generation were selected based on the selection index values. Three selection indices were compared: CCPS, PLS1, and PLS2. CCPS combined crossbred paternal half-sib mean, purebred paternal half-sib mean including the animal itself, purebred full-sib mean including the animal itself, and animal’s own phenotype (females). PLS1 included only purebred information (own performance for females, full-sib mean, and half-sib mean). PLS2 is similar to PLS1 but additional purebred progeny for sires were produced for test purposes. These test progeny were not available for selection. In CCPS the trait under selection was crossbred performance and under PLS1 and PLS2 the trait under selection was purebred performance. In CCPS and PLS2 dams were selected randomly for breeding "test" progeny. Crossbred offspring were produced using a reciprocal mating between the lines. The weighting factors of the index for CCPS are defined in Wei and van der Werf (1994) and the weighting factors of the indices for PLS can be easily derived using general selection index theory and are not shown here.

Because the trait was sex limited all male full-sibs had the same index value, thus the number of males from each full-sib group eligible for selection was limited to one. Mating pairs were randomly assigned, except that mating between full-sibs was avoided to reduce the level of inbreeding in each generation. Generations (total of 15) were discrete and at the end of each generation the population mean for the purebred and crossbred performances and the allele frequencies of the loci were stored. From these values the mean of the F1 population and r_pc were calculated (Falconer and Mackay, 1996; Wei et al., 1991). Simulations were replicated 100 times. Later, the notation PLS refers to both PLS1 and PLS2 simultaneously.

RESULTS AND DISCUSSION

Mean and standard deviation of genetic progress and r_pc under different genetic models are given in Table 1. Under low to moderate degrees of dominance (d=0.5 to 0.75) PLS and CCPS gave similar responses and variances of the response. When dominance was complete, CCPS outperformed PLS, especially, in the early stages of selection. Under overdominance CCPS was clearly better than PLS, however, CCPS had also higher variance than PLS.

Under partial dominance frequency of the favorable allele increased due to selection and the lines became more alike thus r_pc approached 1.00. If dominance is complete, selection cannot distinguish between either homozygotes and heterozygotes leading to continued segregation of both alleles at all loci in both lines even after several generations of selection. Thus the variation in the differences between allele frequencies at a given loci remains relatively high and the loci with large allele frequency differences between lines contribute substantially to the overall measure of r_pc. If the trait is controlled by over-dominant loci the proportion of heterozygotes within selected individuals remains high if CCPS is practiced over several generations leading to decreased and eventually negative r_pc.

Previous calculations by Wei and van der Werf (1994) suggested that CCPS results in faster genetic progress than selection based on pure line information whenever the correlation between the correlation between pure and crossbred performance is less than 0.8. The results presented here are in general agreement with those results but lead to rather different practical interpretation. Firstly, the advantage of CCPS was very small even for values of r_pc well below 0.8, unless overdominance is involved. Also, based on these and additional simulations (results not shown here) the rank of the strategies (CCPS vs PLS) depends on the degree of dominance not the initial differences in allele frequencies between lines. Hence, the rank of CCPS vs PLS is not closely related to the initial level of dominance variation. For example, when both lines had the same initial allele frequencies at all loci but dominance was complete (r_pc=1), r_pc decreased after few generation of selection and CCPS gave slightly faster genetic progress than PLS.

As a conclusion, CCPS will give marginal gains in selection response unless dominance variation is principally due to overdominant gene action, even though there may be a low r_pc and substantial dominance variation. Breeders having evidence of dominance variation within lines have to decide between two strategies: either maximizing the rate of genetic progress by using crossbred information and paying for extra testing of the crossbred animals, or risk losing some genetic response by assuming that the most probable reason for dominance variation is a group of genes with a low to intermediate partial dominance. The latter course would save money from crossbred testing or allow allocating the available housing resources for other purposes.

REFERENCES

Falconer, D.S. and Mackay, T.F.C. (1996) "Introduction to Quantitative Genetics" 4^th ed. Longman, Essex.
Hunton, P. (1990) In ‘Poultry Breeding and Genetics’ pp. 985-1028, editors R. D. Crawford, Elsevier Science Publishing Company Inc., Amsterdam.
Wei, M. and van der Werf, J.H.J. (1994) Anim. Prod. 59:401-413.
Wei, M., van der Werf, J.H.J. and Brascamp, E.W. (1991) J. Anim. Breed.and Genet. 108:262-269.