Fig. 1.

Schematic illustration of the stepwise approach for improving prediction accuracy and maximizing advantages of genomic selection in CIMMYT’s tropical maize breeding program. The strategy proposes the use of algorithms to select individuals from historical data to achieve high predictive performance for new populations and enable advancement into stage 2 yield trials. When historical data are not sufficient to predict the performance of new lines, test-half-predict-half and sparse testing approaches are used