Reconstructing relationship matrices from dense SNP arrays for the prediction of genetic potential in unreplicated multilocation plantings of apple progeny

Prediction of the genetic potential of a candidate cultivar is fundamental to achieving a response to selection in plant breeding. In the early phases of advanced selection plantings, a common strategy to increase selection intensity is field testing of many candidates at multiple locations with incomplete or limited replication among trials. Genetic analyses that incorporate genetic relationships among candidates can be used to improve the prediction accuracy, and indirectly predict performance at locations where a candidate is not planted. Here, we report on the prediction of breeding value for sensory crispness for unreplicated apple candidate progeny and parents established at three locations across the USA (Minnesota, New York State, Washington State) using relationships constructed from (i) pedigree records and (ii) 8K single nucleotide polymorphism (SNP) array data. The correlation between relationship coefficients estimated from either historical pedigree or SNP arrays was 0.76. Estimates of phenotypic variation and heritability were similar among three analyses (i.e., full data using pedigree relationship matrix, reduced data for only individuals with SNP genotypes available using a pedigree-based relationship matrix, or reduced data using a genomic relationship matrix estimated from the SNP arrays). Significant G×E was detected with all analyses, although there were some differences in the estimates of additive genetic correlation among locations from different analyses. This study demonstrates that dense SNP marker arrays may be used to predict breeding values for incompletely or unreplicated candidate cultivars established across multiple locations. This approach will be particularly valuable where the cost or logistics of replication at multiple locations is prohibitive and when pedigree records are incomplete and are not able to fully characterize the diversity of a candidate population.