Frequency of superior yielding segregates from soybean crosses ranked on the basis of best linear unbiased predictions of parents

Plant breeding can be improved by methods that give reliable predictions of crosses which produce the highest frequencies of segregates which are higher yielding than existing genotypes. Best Linear Unbiased Predictions (BLUP) of parents were used by Panter and Allen (1995) to rank 24 good x good soybean (Glycine max [L.] Merr.) crosses for seed yield. Six crosses from this previous study were chosen in order to determine: (i) if the prediction of superior versus inferior crosses by BLUP holds true in advanced generations based on the frequencies of superior yielding lines and the source of the best line; (ii) if discarding lower ranked crosses would have resulted in loss of superior yielding lines; and (iii) the correlations between the genetic variance, genetic relationship of the parents, cross mean yield, and the number of superior yielding segregates. The bases for choosing the six crosses were their yield rank based on BLUP, the genetic relationship between the parents, and the within-cross genetic variance. Fifty to sixty F₇- derived lines per cross were yield tested over a three year period at four locations in incomplete block designs. The least squares means for the individual lines were used to determine the frequency of lines per cross that were one or two standard deviations above the grand mean of all sbc crosses. The two top-ranked crosses as predicted by BLUP accounted for 76% and 86% of the superior segregates that were one or two standard deviations above the grand mean, respectively. The two mid-ranked crosses accounted for 2% and 0% of superior segregates; whereas the two lower ranked crosses accounted for 22% and 14% of superior segregates. The only statistically significant correlation found was between the cross mean yield and the number of superior segregates (0.87, P = 0.02). BLUP accurately predicted the top crosses with the greatest proportion of high yielding segregates. Therefore this method of mean yield prediction appears to be a reliable method of predicting superior soybean crosses.