Multienvironment Validation of the Effectiveness of Phenotypic and Genomewide Selection within Biparental Maize Populations

ABSTRACT Published information is lacking on whether genomewide selection, based on a single tester and a single year of testing, can identify maize ( Zea mays L.) lines that would perform well in multiple subsequent years and with multiple testers. Our objectives were to determine (i) if phenotypic selection or genomewide selection is more predictive of maize performance in future environments and with different testers, (ii) if combining both marker and phenotypic information is advantageous in selection, and (iii) the upward bias in correlations between marker‐predicted values and phenotypic values ( r MP ) when cross‐validation across individuals and across environments is not performed. We evaluated four elite populations, each with 150 or 250 doubled haploid (DH) lines, in 18 environments in the US Corn Belt. The DH lines were genotyped with 3072 single nucleotide polymorphism markers. The r MP values were 0.14 to 0.66 for grain yield and 0.49 to 0.66 for moisture. Phenotypic selection was always as effective as or more effective than genomewide selection. The r MP was lower when different testers were used for the training and test populations. Selection based on marker and phenotypic information was slightly more effective than genomewide selection or phenotypic selection alone. The correlation between predicted and observed performance was higher when cross‐validation across individuals and years was not performed than when cross‐validation across individuals and years was performed. For genomewide selection to be superior to phenotypic selection, the gains must not be measured in terms of the per‐generation response with equal population sizes and selection intensities.