Relationship in the selection for productivity and oil and protein contents in soybean using mixed models

The objective of this work was to evaluate the influence of relationship information for selecting soybean progenies as to their productivity, and oil and protein contents, using mixed models for the prediction of breeding values. Nine hundred F4:6 and 200 F4:7 soybean progenies were evaluated in the seasons 2010/2011 and 2011/2012, respectively. The progenies were obtained from multiple crosses from 57 parents. Data were analyzed using random models (least squares) and mixed models BLUP/REML (best linear unbiased prediction/restricted maximum likelihood). The highest values of predicted gains were obtained by BLUP/REML. The breeding values predicted with the use of BLUP/REML without relationship information were highly correlated with the ones obtained with the random model, and the selected progenies were rather coincident. The inclusion of the relationship matrix resulted in the selection of different progenies and in higher accuracy of breeding values.

Glycine max; BLUP/REML; selection gain; relationship matrix