Modelling interaction from top-cross design data and prediction of F1 hybrid value

Abstract

In breeding for hybrid varieties, top-cross mating designs are widely used to estimate the combining ability of inbred lines as a measure of their potential as parents of commercial hybrids. A multiplicative model is proposed to take into account interaction effects in addition to general combining ability effect (GCA) estimates to predict the values of potential line × line single cross hybrids. The efficiency of this approach was studied on a set of 58 maize inbred lines representing a wide range of germplasm. These lines were crossed to 4 tester lines and hybrids were evaluated for silage yield. The value of line × line hybrid values were predicted according to several models with and without multiplicative terms. The efficiency of the models was estimated by comparing the predicted and the observed values for a sample of 88 line × line reference hybrids. Predicted values were significantly correlated with observed values for every model. For a given set of testers, models which included multiplicative terms were most efficient. A simulation approach confirmed this conclusion for the best hybrids. Thus, this type of model provides a possibility to increase the efficiency of prediction without requiring additional measurements and should be tested for various germplasms. © 1993.