Genomic prediction of sunflower hybrid performance

Predicting single-cross performance is of high importance to improve the efficiency of sunflower (Helianthus annuus L.) hybrid breeding programmes. We used experimental data from inter- and intragroup sunflower hybrids and their parental lines adapted to Central Europe to (i) study the genetic diversity and combining ability and (ii) examine the accuracy to predict hybrid performance based on phenotypic and genomic data. We evaluated 133 intragroup and 104 intergroup hybrids with their parental lines in replicated trials at four environments for grain yield, oil yield and oil content. Furthermore, the parental lines were fingerprinted with 572 AFLP markers. Variance due to specific combining ability was comparable for intergroup and intragroup crosses. This suggested a lack of clearly defined heterotic groups for the sample of studied sunflower lines. Prediction accuracy of hybrid performance based on general combining ability effects was high and could not be increased using genomic selection approaches. For situations where no information on GCA effects of parental lines was available, hybrid prediction based on genomic selection methods was accurate for groups of related lines. For groups of unrelated lines, however, we observed a strong decrease in the prediction accuracy. This suggests that prediction of hybrid performance for crosses based on genetically distant parents remains challenging.