Epistasis in Maize (Zea mays L.): III. Significance in Predictions of Hybrid Performances1

Abstract

The relative significance of epistatic effects and genotype ✕ environmental interaction effects in predictions of hybrid performances was evaluated in 28 sets of single, three‐way, and double crosses. Parental inbred lines originated from two unselected maize (Zea mays L.) varieties and from these two varieties after three cycles of reciprocal recurrent selection. Comparisons between unselected (UNS) and selected (RRS) population means indicate that selection has influenced both the additive and nonadditive genetic contributions to the observed differences.Comparisons among means over sets show significant epistatic effects, which, on the average, cause underestimates of three‐way and double‐cross performances from predictions based on nonparental single cross means. However, these underestimates appear to be inconsequential except for a few unique combinations of lines. Correlations between predicted and observed performances indicate that both the effects of epistatic deviations and the effects of genotype ✕ environmental interactions cause significant biases in hybrid predictions. Predictions of yield and ear number based on data from a single environment show somewhat greater biases from genotype ✕ environmental interactions than from epistatic effects. However, when predictions are based on data from three environments, biases from the two types of effects differ little in magnitudes.Except for those unique, apparently infrequent combinations of lines very strongly influenced by epistatic effects, the results indicate that standard single‐cross procedures for predicting three.way and double crosses need not be replaced by more complex procedures. Data from a single environment provide poor discrimination among three‐way or double crosses and somewhat better discrimination among single crosses. It is suggested that preliminary selections for three‐way and double‐cross hybrids are most efficiently performed by utilizing predictions based on nonparental single‐cross means from a few environments. Final testing of reduced numbers of selected hybrids requires several environments for adequate evaluation. Likewise, ascertainment of commercially acceptable unique single‐cross combinations requires several environments.