Search

The NCII design (North Carolina mating design II) has been widely applied in studies of combining ability and heterosis. The objective of our research was to estimate how different base populations, sample sizes, testcross numbers and heritability influence QTL analyses of combining ability and heterosis. A series of Monte Carlo simulation experiments with QTL mapping were then conducted for the base population performance, testcross population phenotypic values and the general combining ability (GCA), specific combining ability (SCA) and Hmp (midparental heterosis) datasets. The results indicated that: (i) increasing the number of testers did not necessarily enhance the QTL detection power for GCA, but it was significantly related to the QTL effect. (ii) The QTLs identified in the base population may be different from those from GCA dataset. Similar phenomena can be seen from QTL detected in SCA and Hmp datasets. (iii) The QTL detection power for GCA ranked in the order of DH(RIL) based > F based > BC based NCII design, when the heritability was low. The recombinant inbred lines (RILs) (or DHs) allows more recombination and offers higher mapping resolution than other populations. Further, their testcross progeny can be repeatedly generated and phenotyped. Thus, RIL based (or DH based) NCII design was highly recommend for combining ability QTL analysis. Our results expect to facilitate selecting elite parental lines with high combining ability and for geneticists to research the genetic basis of combining ability. [ABSTRACT FROM AUTHOR]

Selection index (SI) theory has been applied to predict the net genetic merit, select parents for the next cycle, and maximize the selection response in plant breeding. However, up to now, SI has not been applied to predict unobserved hybrid performance. In hybrid breeding, it is impossible to test every cross, and accurate prediction can help breeders greatly reduce the experimental cost. Traditional genomic selection (GS) targets single-trait prediction, and useful information about other related traits is ignored. With the data set of 575 rice hybrids subjected to the North Carolina mating design II, this information was utilized to develop a linear SI-based GS method to predict rice hybrid for a more accurate and comprehensive selection. Cross-validation results showed that genetic information of a low-heritability target trait such as grain yield could be greatly aggregated from auxiliary traits using SI. When SI was used for directly predicting the target trait, the SI-direct prediction underperformed the traditional genomic prediction in most cases. However, when the SI-direct prediction was combined with the traditional genomic prediction using a suitable weight ω inferred from the training data set, significantly higher accuracy could be obtained. This method was called as SI-assisted prediction. It provided a promising prediction means for breeding application, which used the phenotypes of auxiliary traits only in the training data set. Additionally, it was found that SI-assisted accuracy increased as the genetic correlation between auxiliary traits and the target trait increased, and high heritability of auxiliary traits could also improve the prediction performance. [ABSTRACT FROM AUTHOR]