How can we best use DNA data in the selection of cattle?

Some traits are controlled by single genes, such as many genetic abnormalities, and some cattlemen already use DNA tests for these genes. However, most economically important traits are controlled by many genes and influenced by the environment. Recently panels of 50,000 genetic markers called SNPs have become available and they provide an opportunity to select for all the traits we seek to improve. The first step in using these SNPs is to estimate the effect of each SNP on all the important traits. To do this requires a ‘discovery population ’ of cattle that have been genotyped for the SNPs and measured for the traits. From this population a prediction equation is derived that predicts breeding value or progeny difference for a trait, say marbling, from the 50,000 SNP genotypes (breeding values are just twice the progeny difference). This prediction equation can then be applied to animals that have SNP genotypes but no phenotypic information to calculate a predicted breeding value that is sometimes called a molecular breeding value (MBV). Usually animals will have some other source of information about their breeding value (e.g. pedigree and ultrasound scans) that are used at present to calculate a traditional estimated breeding value (EBV) or EPD. Therefore the MBV should be combined with the traditional EBV to give a prediction of breeding value that is more accurate than either source of information alone. This new estimate has been called a genomic EBV (GEBV). Obtaining a prediction equation that accurately predicts breeding value from SNP genotypes has proven difficult in beef cattle. We have often found that a prediction that works in one breed or herd does not work in other breeds and herds. Therefore we need to estimate the prediction equation from very large discovery populations which include several breeds and the test or validate the prediction equations across large populations also comprising several breeds