Abstract:

The breeding values of eight Paulownia species were predicted by Best Linear Unbiased Prediction (BLUP) and Best Linear Prediction (BLP), Results showed that there were no significant effects on predicted results of breeding values based on BLUP whether with numerator relationship matrix (NRM) constructed by RFLP of cpDNA or not. If too much data available, breeding values could be predicted utilizing plot means values instead of individual datum or by the BLP instead of BLUP. Correction between the predicted values and the true genetic values was closely related to the variance among the predictions or the error variance of predictions, so the former could be used to indicate the precision of breeding values instead of later two coefficients. By compared variance and error variance among predicted breeding values based on provenance trails of P. tomentosa, BLP was a more reliable method to rank candidates than that of the least squares estimates. However, if repetitions were more than five, and individuals in every plots were over six, the precision of these two methods for breeding values prediction had little difference.

Key words: Paulownia, Breeding value, Best linear unbiased prediction, Best linear prediction