Since its success in China in the 1970s, hybrid rice technology has spread to other tropical countries such as India, Vietnam, Bangladesh, and the Philippines. Compared with inbred cultivars, the development of hybrid rice is more costly because it entails identification of parental lines that produce hybrids with superior yield. To reduce the number of crosses and the time consumed in field evaluation, an efficient prediction method for hybrid performance would be useful.
Midparent value (MPV), which is the average of both parents of a hybrid, is a classical method of predicting hybrid performance. This method, however, proved to be less efficient in predicting for traits with low heritability. Genetic distances derived from molecular markers were reported to have a certain significant correlation with hybrid yield performance, but this method is expensive. Notably, the best linear unbiased prediction (BLUP) procedure has been successfully applied to predict single-cross performance of maize and soybean (Bernardo 1994, Panter and Allen 1995).