Genomic Selection Using Maize Ex-PVP Germplasm for the Prediction of N-Use Traits.

Nitrogen use efficiency (NUE) in maize (Zea mays L.) is an important trait to maximize yield with minimal input of N fertilizer. Current breeding and biotechnological approaches have failed to develop a maize hybrid with optimal NUE. Genomic selection (GS) for N-use traits may speed up the breeding cycle of research programs targeting for improved NUE in maize. The objectives of this research were to evaluate the prediction accuracy of GS for different N-use traits in maize hybrids using G-BLUP and evaluate the impact of training population sizes and training composition for effective application of genomic prediction on NUE breeding programs in maize. Eighty-six ex-PVP inbreds (33 stiff stalk synthetic, SSS and 53 non-stiff stalk, NSS) were genotyped with 26,769 single-nucleotide polymorphism (SNPs) and 259 single-cross maize hybrids were grown in eight environments during the years of 2011 and 2015 at two N fertilizer rates (0 and 252 kg N ha^-1), and three replications per environment. Across different traits and training set sizes prediction accuracy ranged from 16 to 42% if either none or both of the hybrid parents in the training set were in the validation set. Harvest index exhibited the greatest prediction accuracy, ranging from 16 to 85% at high N and 30 to 97% at low N. The higher prediction accuracy at low N versus high N could have been due to the less complex genetic architecture of the NUE traits under N stress. The N-use trait or traits with high prediction accuracy may be integrated into maker-assisted breeding strategies for improved NUE.