Simulating past selection for molecular breeding program evaluation

Abstract

Domestic animals have been selected by their keepers for thousands of years, by choosing more desirable animals to breed the next generation. Marker assisted selection (MAS) is a relatively new concept, whereby parents may be selected on the basis of their 'molecular breeding values,' possibly in conjunction with pedigree information. This is important in the dairy industry, where generation intervals can be up to seven years, spanning the time from planned matings to completion of bulls' daughters' first lactations. Young bulls are initially selected for inclusion in a progeny test program (at a cost of at least A$20,000 each) on the merit of their relatives, but only about three percent of these young bulls are ultimately widely used by artificial insemination. More accurate initial selection of these young bulls using MAS could reduce progeny testing costs and/or increase rates of genetic gain in the dairy industry. A number of programs are available for identifying potential or putative markers, such as QTL Express (Seaton, et al., 2002), QxPaK (Pérez-Enciso & Misztal, 2004) and ASReml (Gilmour et al., 2001), but validation and comparison of different methods is difficult, partly because suitable simulated datasets with QTL of known effect are unavailable. A simulated dataset which incorporates selection could test the efficiency of QTL detection by these programs. Therefore we produced a method of simulating datasets for animals where the trait they expressed (Australian Profit Ranking (APR) in this case) was based on molecular breeding values, reflected past selection and environmental error, and where the size and location of QTL effects was known but concealed. One of these datasets was used to verify analysis methods used for MAS of dairy bulls. This method could also be applied to other animals that were selected on genetic merit, and for simulation of power in detection and application of DNA markers in MAS.