Genomic prediction for apple breeding

Genomic prediction for apple breeding

Genomic prediction relies on the use of statistical models enabling to identify interesting individuals stemming from crosses at their youngest age, using only genotypic data. The models are built using genotypic and phenotypic data obtained on a so called training or reference population. Our team, together with the VaDiPom team and the Experimental Horticultural unit of INRAE in Angers (UE Horti), is part of a European consortium that created such a reference population for apple and planted it in a multisite orchard trial consisting of six locations, in Belgium, Spain, France, Italy, Poland, and Switzerland, in order to generate the required data for this approach. This represents a unique network to date for apple.

The apple reference population consists in 534 genotypes, distributed in 269 mostly old accessions and 265 progeny from 27 biparental combinations, representing the diversity in cultivated apple and 'elite' material from current European breeding programs, respectively. A high-density genome-wide dataset of more than 300,000 SNPs was produced for these genotypes as a combined output of two SNP arrays of different densities using marker imputation. Environmental adaptation traits, floral emergence and harvest dates, were studied in the six locations where the population was planted in the first year of harvest. Predictive abilities of 0.57 and 0.75 were estimated for floral emergence and harvest dates, respectively. These results confirmed the suitability of the reference population for genomics-assisted breeding in apple. Its implementation in six locations, together with the coordinated phenotyping of numerous traits of agronomical interest, including yield, fruit size and quality and phenology, will enable to study genotype X environment interactions (GxE) and the impact of multi-trait prediction model to increase prediction abilities and thus apple breeding program efficiency.



Besides, we study the possibility of using genomic prediction in the framework of favorable allele transfer from genetic resources to elite material. Whereas genetic resources contain a large diversity, with more than 10000 described cultivars worldwide, commercial apple cultivation is indeed dominated today by a very small number of cultivars, and breeding programs rely strongly on the use of genotypes related to these. Enlarging diversity through a more efficient use of genetic resources seems necessary to reduce the vulnerability associated with this genetic uniformity. In this context, we are studying predictive abilities that can be obtained when predicting hybrids between genetic resources and elite cultivars or progenitors, building the prediction model either with data from the reference population or with data previously obtained on genetic resources on one hand, and on elite material on the other hand. We are then applying the models to first generation hybrids that were genotyped and phenotyped. To go beyond the first generation of crosses between genetic resources and elite material, we use simulation approaches.

Associated publications:

Jung M, Roth M, Aranzana MJ, A. Auwerkerken, M. Bink, C. Denancé, C. Dujak, C.-E. Durel, C. Font i Forcada, C.M. Cantín, W. Guerra, N. Howard, M. Lewandowski, M. Ordidge, M. Rymenants, N. Sanin, B. Studer, E. Zurawicz, F. Laurens, A. Patocchi, H. Muranty (2020) The apple REFPOP—a reference population for genomics-assisted breeding in apple. Horticulture Research 7:189. doi: 10.1038/s41438-020-00408-8

Modification date : 20 December 2023 | Publication date : 06 September 2021 | Redactor : ResPom