Strategies for applying marker assisted selection in nucleus breeding schemes in dairy cattle

The general goal of this thesis was to use simulation to examine practical issues for marker assisted selection (MAS) of dairy cattle. Markers were used to select within-family the bulls to enter progeny testing from a nucleus herd. The first study evaluated the effects of considering a confidence interval of the position for a quantitative trait loci (QTL) versus only the probable genotype at the predicted site of the QTL. The location of the QTL was estimated by interval mapping with a granddaughter design. Accounting for the confidence interval increased the response in all scenarios. The average true breeding value (TBV) of the selected bulls was increased 2.60% when the confidence interval was used, versus 2.00% when only the predicted location was considered. No differences were observed with respect to how the confidence interval was estimated.; The second study compared strategies for repeated application of QTL detection and MAS. Twenty QTL and 300 markers were randomly distributed across 30 chromosomes. A daughter design was used, every generation, to determine the associations between marker and QTL alleles. Maximum response was achieved by strategies that selected upon several markers flanking multiple QTL. The mean TBV of selected bulls was increased by up to 12% when multiple loci were considered, versus ≤7% when only the best marker was used.; The third study examined MAS when the selection goal included two traits. Trait 1 had an economic weight and heritability three times greater than trait 2. Multiple trait MAS was compared to applying MAS for trait 1 only and conventional selection alone. Multiple trait MAS decreased response for trait 1 relative to both single trait MAS or conventional selection. However, response for trait 2 increased to a greater degree and, therefore, response for the final index was greater. This result was consistent whether the traits were positively or negatively correlated.; The final study examined how different assumptions about the underlying genetic model affected the long-term response to MAS. Models differed in terms of mutation rate and distributions of allelic effects and frequencies. The use of MAS was beneficial regardless of the genetic model.