BEST LINEAR UNBIASED PREDICTION OF MAIZE SINGLE-CROSS PERFORMANCE

In preliminary studies, best linear unbiased prediction (BLUP) has bee n found useful for identifying high-yielding maize (Zea mays L.) singl e crosses prior to held evaluation. In this study, the effectiveness o f BLUP for large-scale prediction of yield, moisture, stalk lodging, a nd root lodging was investigated. Multilocation data from 1990 to 1994 were obtained from the hybrid testing program of Limagrain Genetics. For each of 16 heterotic patterns, the performance of m untested singl e crosses was predicted from the performance of n tested single crosse s as y(M) = C-MP C-PP(-1) y(P), where y(M) = m x 1 vector of predicted performance of the untested single crosses; C-MP = m x n matrix of ge netic covariances between the untested single crosses and the tested s ingle crosses; C-PP = n x n phenotypic covariance matrix among the tes ted single crosses; and y(P) = n x 1 vector of average performance of the tested single crosses, corrected for yield trial effects. Correlat ions between predicted and observed performance were obtained with a d elete-one cross-validation procedure. For heterotic patterns with larg e (>100) numbers of tested single crosses, the correlations ranged fro m 0.426 to 0.762 for yield, 0.754 to 0.933 for moisture, 0.300 to 0.73 9 for stalk lodging, and 0.164 to 0.532 for root lodging. The correlat ions, especially for lodging traits, increased as larger numbers of te sted single crosses were available. The results in this study were obt ained from large and diverse data sets (600 inbreds, 15 183 data point s, and 4099 tested single crosses across 16 heterotic patterns) and pr ovide strong evidence that BLUP is useful for routine identification o f superior single crosses prior to field testing.