Lessons From Genomewide Selection in Agronomic Crops

Genomewide selection (or genomic selection) allows marker-based selection without QTL mapping. In genomewide selection, equations that predict genotypic value are first developed from phenotypic and marker data in a training population. The prediction equations are then used to assess genotypic values in a test population that has been genotyped, but not phenotyped. Genomewide selection is therefore predicated on genotyping being cheaper and quicker than phenotyping. Key lessons from applying genomewide selection in agronomic crops will be reviewed. First, genomewide selection is most advantageous when heritability is high in the training population, but is low or zero in the test population. Several examples of how such situation can be achieved will be presented. Second, predictions are usually most accurate with a simple model that assumes that each marker accounts for an equal proportion of the total genetic variance. Third, predictions are usually most accurate if epistasis is assumed absent. Fourth, traits differ in their prediction accuracies even when the marker density, population size, and heritability are kept constant. Empirical data are therefore needed to determine which traits are the most predictable and which traits are not. Fifth, for finding marker-trait associations, models that incorporate genomewide background effects are superior to composite interval mapping and to the QK model used for association mapping. The presentation will conclude by discussing future needs in applying genomewide selection in plants.