Relative Efficiency of Genomewide Selection for Testcross Performance of Doubled Haploid Lines in a Maize Breeding Program

Abstract

ABSTRACTIn a simple scheme for genomewide selection in maize (Zea mays L.), a subset of n out of N doubled haploid (DH) lines from a biparental cross are genotyped and phenotyped in testcross combination. The n lines are then used as a training population to predict the performance of the remaining N − n lines. Our objectives were to (i) determine whether or not this scheme for genomewide selection is more efficient than phenotypic selection under a fixed budget and (ii) give recommendations for implementing this genomewide selection scheme during maize line development. From empirical data across multiple populations, testers, years, and locations, we assessed the observed and expected gains from phenotypic selection and genomewide selection under different budget scenarios. Total budgets ranged from $12,000 to $22,000, and standard costs per DH line were $90 for field trials and $15 for genotyping. The relative efficiency of genomewide selection (compared with phenotypic selection) depended on the (i) effectiveness of phenotypic selection and of genomewide selection for the trait in a given cross, (ii) sizes of the training and test populations that could be accommodated within a given budget, and (iii) number of selected lines. In general, the correlation between marker‐predicted values and phenotypic values had to exceed 0.50 for the genomewide selection scheme to be more efficient than phenotypic selection. Our results suggested that for the genomewide selection scheme to be effective, a training population of 60 to 80 lines and a test population of at least 100 lines are needed per biparental cross.