In plant breeding programs for line varieties and hybrid components, superior lines are selected from a breeding pool as parental lines for the next breeding cycle. However, not all possible crosses between the parental lines can be evaluated in the field due to limited resources. It is therefore more efficient to preselect the most promising parental combinations based on genotypic values. Possible progenies of a cross can be characterized by distribution with parameters mean and the standard deviation. For a cross with large mean and standard deviation it is more likely to and superior progeny. Recombination of elite breeding material often results in crosses with similar mean, therefore the standard deviation can help to discriminate among the crosses.Predicting the genetic segregation variance of crosses based on genomic data is currently of great interest in the animal and plant community. This is underlined by a range of recent publications in this field. The knowledge of expected mean and segregation variance is helpful to use selection criteria like the concept of usefulness or superior progeny value. Both selection strategies involve the distribution parameters mean and segregation variance to assess the cross value. Parameter mean can be predicted with simple methods, while predictors associated with the segregation variance were hard to find and made the selection strategies difficult to employ. Distance measures based on phenotypic, genotypic or pedigree data were insensitively studied but were not robustly associated with the segregation variance. The combination of phenotypic and genotypic data, lay the foundation for the recently published methods. At present, simulation approaches and analytical formulas were published to estimate mean and segregation variance based on marker effects that are predicted with genome-wide prediction models. Therefore, new approaches are an extension of genomic prediction that can be obtained with less effort since the method is gaining ground as a tool in breeding programs where genotypic and phenotypic data is routinely available.The first study presents a resource-effcient tool for breeders to select parental lines within a line or hybrid breeding program to distinguish between the most promising crosses that could be made. The estimation is based on marker effects that are predicted with genome-wide prediction models and accounts for the expected gametic disequilibrium between two loci. The derived formulas can be used for typical mating systems like single seed descent and doubled haploid lines, and also consider several generations of intermating before inbred line derivation. A published maize data set was tested and compared with the simulation approach PopVar. The analytical results for means and variances are highly correlated to simulation results. In times of big data management the formulas have a promising speed advantage. At that time, the prediction of mean and segregation variance and application of usefulness and superior progeny value has been tested with simulated data and mapping populations. However, breeders´ data sets represent a major field of application for cross prediction. In the second study, the practical applicability of an analytical approach for cross prediction based on genome-wide marker effects in a real-life barley data set from an ongoing resistance breeding project. The presented approach is fast and convenient to use, and suffciently accurate to identify the 50 % best crosses from the field trial. The new methods are promising for increasing response to selection of line and hybrid breeding programs by extending genomic prediction approaches. The application can support the selection of crossing partners, optimizing or reducing resource use for phenotyping and maintaining genetic diversity in breeding programs.
Verknüpfung zu Publikationen oder weiteren Datensätzen
