Search

Marker-assisted selection schemes that utilize information about quantitative trait loci to preselect progeny test bulls within a family are the most practical application of quantitative trait loci results in the short-term. Technical difficulties exist for across-family marker-assisted selection using BLUP procedures. Two within-family marker-assisted selection schemes were evaluated genetically and economically using stochastic simulation for a locus that explained 5% of phenotypic variance. The genetic and economic impacts of variation in the number of offspring per bull-dam were evaluated. The top down marker-assisted selection scheme identifies sires that are heterozygous for the locus based on the granddaughter design and uses the quantitative trait locus information in the preselection of grandsons entering progeny testing. The bottom up marker-assisted selection scheme identifies sires heterozygous for a quantitative trait locus based on the daughter design and uses the information in the preselection of sons entering progeny testing. The top down scheme with one progeny per bull-dam reduced the rate of genetic gain compared with that from a breeding scheme that ignored knowledge of the quantitative trait locus. The top down scheme with reproductive performance of 3 or 40 progeny per bull-dam increased genetic gain by 1 to 2%. The bottom up scheme increased the rate of genetic gain by 1.5, 3.5, and 5% for 1, 3, and 40 progeny per bull-dam, respectively. When the top down scheme was used on the maternal path and the bottom up scheme on the paternal path, increases were 9% with 40 progeny per bull-dam. The use of reproductive technologies on bull-dams is imperative to prevent gains from marker-assisted selection being eroded by the loss in polygenic selection differential that results when more bull-dams are required to enable preselection of sons using markers.

Results from quantitative trait loci studies cannot be readily implemented into breeding schemes through marker assisted selection because of uncertainty about whether the quantitative trait loci identified are real and whether the identified quantitative trait loci are segregating in the breeding population. The present paper outlines and discusses strategies to reduce uncertainty in the results from quantitative trait loci studies. One strategy to confirm results from quantitative trait loci studies is to combine P-values from many quantitative trait loci experiments, while another is to establish a confirmation study. The power of a confirmation study must be high to ensure that the postulated quantitative trait loci can be verified. In the calculation of the experimental power, there are many issues that have to be addressed: size of the quantitative trait loci to be detected, significance level required, experimental design and expected heterozygosity for the design. To ensure marker assisted selection can be quickly implemented once quantitative trait loci are confirmed, DNA samples should be retained from daughters, and the sires and dams of elite sires.

The effect of inaccurate estimates of variance and of the location of the quantitative trait locus on the genetic response to marker-assisted selection was studied by simulation of an adult multiple ovulation and embryo transfer nucleus breeding scheme. Two genetic models were simulated for the quantitative trait locus: a total of 10 alleles or 2 distinct alleles per base parent. For both models, the locus explained either 5 or 10% of phenotypic variance. A polygenic component was simulated, and the two genetic components were summed to 35% heritability for a trait measured on females. Overestimation of variance of the quantitative trait locus had minimal effect on genetic gain for marker-assisted selection over the short term, but decreased long-term response. The long-term loss was reduced when variance of the quantitative trait locus was reestimated after four generations of marker-assisted selection. Selection for favorable alleles at a nonexistent quantitative trait locus resulted in first generation losses of 3 and 7% for postulated quantitative trait loci, explaining 5 and 10% of variance, respectively. The larger the degree of error in location, the larger was the genetic loss compared with the correct location scenario. For the largest simulated location error of 15 cM, genetic superiority of marker-assisted selection was reduced by 80% in the first generation. We concluded that studies should be undertaken to verify estimates of quantitative trait locus and location to make optimal use of marker-assisted selection.

Five traits of major economic importance in the New Zealand dairy industry are milk volume, milk fat, milk protein, live weight, and survival. This study evaluated the impact of live weight as a trait in the selection objectives for the New Zealand dairy industry. Live weight of the lactating cow is an important measure because it reflects feeding costs via maintenance feed and salvage values of cows to be culled. In addition, selection responses were evaluated for differing relative economic values for milk protein and milk fat, and selection indexes that included or excluded phenotypic and genotypic correlations between traits were compared. Inclusion of live weight, with a negative economic value in a four-trait selection index with milk, milk fat, and protein resulted in higher economic response. Protein response to selection was not more than 2% when the relative economic value for the ratio of protein to milk fat exceeded 5:1 in a two-trait model; however, milk fat response decreased by over 10%. When a negative relative economic value was assigned to milk fat, economic returns were lower because of lower milk fat responses and the lack of higher protein responses compared with the same ratio for relative economic value but a positive weight for milk fat. Accounting for phenotypic and genetic correlations in deriving selection index weight improved economic response 5%.

Twenty Dutch Holstein-Friesian families, with a total of 715 sires, were evaluated in a granddaughter experiment design for marker-QTL associations. Five traits-milk, fat and protein yield and fat and protein percent-were analyzed. Across-family analysis was undertaken using multimarker regression principles. One and two QTL models were fitted. Critical values for the test statistic were calculated empirically by permuting the data. Individual trait distributions of permuted test statistics differed and, thus distributions, had to be calculated for each trait. Experimentwise critical values, which account for evaluating marker-QTL associations on all 29 autosomal-bovine chromosomes and for five traits, were calculated. A QTL for protein percent was identified in one, and two QTL models and was significant at the 1 and 2% level, respectively. Extending the multimarker regression approach to an analysis including two QTL was limited by families not being informative at all markers, which resulted in singularity. Below average heterozygosity for the first and last marker lowered information content for the first and last marker bracket. Highly informative markers at the ends of the mapped chromosome would overcome the decrease in information content in the first and last marker bracket and singularity for the two QTL model.