Sampling problems with nested mating designs for estimating epistatic variance components in testcrosses of maize

Testcross progenies from two mating designs were developed within different maize (Zea mays L.) populations. Both populations had similar origin hilt one was bred two cycles for silage yield and the other was bred two cycles for grain yield. In the S-1 design, three plants per S-1 family were crossed to the tester while in the nested design (NC I) three plants per full-sib family were crossed to the tester. Such mating designs are well suited for studying the genetic variance among testcross progenies, which can be partitioned into additive and additive x additive epistatic variances. Trigenic epistasis also can be considered in the nested design. S-1 design was replicated in two years. A direct estimation of genetic variance components leads to inconsistencies among the two populations, with some negative estimates and with a surprisingly high apparent contribution of epistatic variances for some traits. The introduction of a parameter of non-random sampling of the plants from a family crossed to the tester (i.e., a negative correlation among such plants), allows more consistent and positive estimates of genetic variance components and decreases the amount of epistatic components. Plant height and ear height were controlled mainly by additive effects in both populations, whereas epistatic variance was more important for earliness traits such as silking date and grain moisture. In the population studied for forage yield and in two years, no epistatic variance was detected, but a significant additive x year variance. The population studied for grain yield tends to show a significant epistatic variance component for this trait.