Genome Wide Selection for Grain Yield of Corn Hybrids Using the Gblup MODEL.

The main objective of a genetic breeding program is to generate individuals more productive than those existent. However, in the case of corn breeding, this identification demands a large number of crosses between inbred lines. In most cases, only a part of these crosses is performed, making it difficult to identify these superior hybrids. Therefore, genomic prediction has become an alternative tool that uses the information derived from molecular markers for the selection of the best combinations aiming at a more efficient breeding. In this work, the objective was to perform genome wide selection using a set of Darts-seq markers and the GBLUP model with dominance for grain productivity of corn hybrids evaluated in different years and locations. A genotyping was conducted with Darts-seq markers on 447 inbred lines derived from a germplasm bank of a private corn seed company. From the crosses of these inbred lines, 838 simple hybrids were obtained, evaluated at six locations in the winter season of 2013, and 797 simple hybrids, evaluated at four locations in the summer season of 2013/2014. For the prediction, we used the GBLUP model with dominance and cross validation with for levels of unbalancing – 10, 20, 30 and 50% and with 100 and 50% of the set of generated markers. The common hybrids from both seasons were used to calculate the correlations. The magnitude of the correlations between predicted and observed hybrids ranged from 0.82 to 0.89 in the winter season and from 0.56 to 0.76 in the summer season. The coincidences between the VGGs of the summer and winter seasons, in terms of disposal, show another dimension of the methodology. This result demonstrates the possibility of using this technique in breeding programs for disposing of non-promising genotypes. The GBLUP method was capable of generating high correlations between predicted and observed hybrids, even with high levels of unbalance and in different locations and years.