Sorghum [Sorghum bicolor (L.) Moench] is a very important crop in the arid and semi-arid tropics of India and African subcontinent. In the process of release of new cultivars using multi-location data major emphasis is being given on the superiority of the new cultivars over the ruling cultivars, while very less importance is being given on the genotype 9 environment interaction (GEI). In the present study, performance of ten Indian hybrids over 12 locations across the rainy seasons of 2008 and 2009 was investigated using GGE biplot analysis. Location attributed higher proportion of the variation in the data (59.3-89.9%), while genotype contributed only 3.9-16.8% of total variation. Genotype 9 location interaction contributed 5.8-25.7% of total variation. We could identify superior hybrids for grain yield, fodder yield and for harvest index using biplot graphical approach effectively. Majority of the testing locations were highly correlated. 'Which-wonwhere' study partitioned the testing locations into three mega-environments: first with eight locations with SPH 1606/1609 as the winning genotypes; second megaenvironment encompassed three locations with SPH 1596 as the winning genotype, and last mega-environment represented by only one location with SPH 1603 as the winning genotype. This clearly indicates that though the testing is being conducted in many locations, similar conclusions can be drawn from one or two representatives of each mega-environment. We did not observe any correlation of these mega-environments to their geographical locations. Existence of extensive crossover GEI clearly suggests that efforts are necessary to identify location-specific genotypes over multi-year and -location data for release of hybrids and varieties rather focusing on overall performance of the entries.

Prediction of single-cross performance has been a major goal of plant breeders since the beginning of hybrid breeding. Recently, genomic prediction has shown to be a promising approach, but only limited studies have examined the accuracy of predicting single-cross performance. Moreover, no studies have examined the potential of predicting single crosses among random inbreds derived from a series of biparental families, which resembles the structure of germplasm comprising the initial stages of a hybrid maize breeding pipeline. The main objectives of this study were to evaluate the potential of genomic prediction for identifying superior single crosses early in the hybrid breeding pipeline and optimize its application. To accomplish these objectives, we designed and analyzed a novel population of single crosses representing the Iowa Stiff Stalk synthetic/non-Stiff Stalk heterotic pattern commonly used in the development of North American commercial maize hybrids. The performance of single crosses was predicted using parental combining ability and covariance among single crosses. Prediction accuracies were estimated using cross-validation and ranged from 0.28 to 0.77 for grain yield, 0.53 to 0.91 for plant height, and 0.49 to 0.94 for staygreen, depending on the number of tested parents of the single cross and genomic prediction method used. The genomic estimated general and specific combining abilities showed an advantage over genomic covariances among single crosses when one or both parents of the single cross were untested. Overall, our results suggest that genomic prediction of single crosses in the early stages of a hybrid breeding pipeline holds great potential to redesign hybrid breeding and increase its efficiency.

No abstract

Prediction of single‐cross performance in a hybrid breeding program is extremely important because it is not feasible to evaluate all parental combinations. Recent simulation and field studies have shown great promise of genomic prediction of single‐cross performance. These previous studies, however, have primarily focused on parametric genomic prediction models. This study tested three nonparametric models—reproducing kernel Hilbert spaces, support vector regression, and neural networks—for prediction of early‐stage single crosses. Two separate datasets, consisting of 481 and 312 single crosses, were used to evaluate models. Single crosses were made by randomly crossing inbred progenies between heterotic groups. Genomic prediction models were trained to directly predict single‐cross performance, or to predict general combining ability (GCA) of inbred parents and specific combining abilities (SCA) of single crosses between them. Using cross‐validation, genomic predictions were compared with predictions using phenotypes of single crosses with a common parent (common‐parent single crosses), as well as phenotypic estimates of GCA. Of these three options for predicting single‐cross performance, genomic prediction resulted in the highest correlation between observed and predicted values. Predictive abilities of parametric and nonparametric genomic prediction models were nearly identical. All genomic prediction models displayed good ability to predict GCA effects, but none could predict SCA effects. Our results suggest that nonparametric models do not provide an advantage over parametric models for prediction of early‐stage, single‐cross performance using modestly sized training populations like those used here.

No abstract