Genomewide selection of parental inbreds: Classes of loci and virtual biparental populations

Having information on the distribution of progeny in a biparental cross—before the cross is made—would be helpful. My objectives were (i) to determine if genomewide markers can be used to accurately predict genotypic values at different classes of loci between two parents (P1 and P2) and (ii) to propose and analyze the use of virtual populations for choosing parents in inbred development. Prior estimates of genomewide marker effects (28,626 single nucleotide polymorphism loci) for silking date and kernel protein among 284 maize (Zea mays L.) inbreds were used. Four classes of loci were determined by which parents had the favorable allele: both P1 and P2; P1 only; P2 only; and neither parent. Simulations indicated that when genomewide predictions for the trait as a whole were accurate, predictions of genotypic value at each class of loci were also accurate. The maize data, however, indicated limited practical value of the classes of loci in choosing P1 and P2. Virtual populations of 1000 doubled haploids were simulated for each of 45 P1 × P2 crosses to predict the population mean, variance (V_DH), and mean of the best 10% of doubled haploids (U_0.10). The predicted mean, predicted V_DH, and predicted U_0.10 for silking date and protein concentration behaved as expected from quantitative genetics theory and were consistent with the genetic backgrounds of the maize inbreds. The usefulness of virtual populations for choosing parents needs to be validated either by empirical experiments or by their routine use in a breeding program.