Plant & Animal Genomes XVII Conference Plant & Animal Genomes XVII ConferenceJanuary 10-14, 2009
Town & Country Convention Center
San Diego, CA
J. D. Nkrumah , B. W. Woodward , S. W. Bauck
Recent advancements in molecular genetics, especially in the sequencing of the genomes of several domestic animals, have resulted in the generation of large amounts of genomics data. This has enabled the development of medium to high-density panels of SNP markers for applications in the characterization of the molecular basis of variation in several economically important traits. Traditionally, the information from these markers has been targeted at estimation of genetic merit for identifying superior individuals early in the production cycle. The potential exists to develop impactful and cost-effective panels of markers for applications in the management of performance and carcass outcomes in the feedlot. Such predictive models could be implemented through innovative data mining approaches as well as traditional linear models. The goal of data mining is to model the training data precisely while maximizing its generalization to making predictions on unseen data. In the current study, we divided SNP genotype and phenotype records on approximately 4,500 feedlot cattle into training, validation, and prediction sets. We then evaluated the potential of using different data mining methods such as linear regression, logistic regression, decision trees, and artificial neural networks to predict feedlot carcass outcomes. The results of such studies should provide significant insights on the potential application of DNA information in the feedlot.