Genomic prediction-aided incorporation of genetic resources into elite breeding: lessons from a collaborative multiparental design in flint maize.

Key message: A public private cooperative mating design between elite maize inbred lines and diversity donors shows that genomic prediction holds great promise to improve the use of genetic resources. Genetic diversity is essential for plant breeding, enabling long-term gains and adaptation to climate change and new agronomical practices. Breeders can access diverse genetic resources to enhance elite germplasm and introduce new favorable variations. The limited performance of genetic resources may hamper their use. To overcome this, a bridging population can be implemented to evaluate and select progenies from crosses between diversity donors and elite lines before their introduction in breeding programs. The choice of such crosses can be dealt with the usefulness criterion (UC), which determines its ability to produce transgressive individuals. This paper investigates the use of genome-wide marker effects to predict (i) the performance of individuals derived from crosses between donors and elite lines and (ii) the UC of crosses not observed yet. It also compares donor introduction strategies based on the UC or the H criterion, which considers the genome-wide donor-elite complementarity. We used a flint maize collaborative multi-parental BC1-S2 population, consisting in materials from six breeding companies and one public institute crossed to different donors. The 20 crosses had contrasted means and genetic variances, and most of them presented transgressive individuals above the elite parent. Results emphasize the importance of half-siblings derived from the elite line parent of the predicted cross to efficiently predict progeny performances or the UC. They also showed that using the H criterion appears promising to select iteratively donors that best complement initial elite materials. The paper concludes with guidelines for implementing a bridging population using genome-wide marker-based predictions.