Genomic prediction enables selection for improved fruit ripening traits in northern highbush blueberry ( Vaccinium corymbosum L.) breeding populations

Blueberry (Vaccinium corymbosum L.) breeding is a time-consuming, complex, and labor-intensive process due to the long-lived woody perennial nature of blueberries and their autotetraploid (2n = 4x = 48) genome. Blueberry breeders have recently adopted and implemented molecular tools for enhancing breeding efficiency and accuracy. Genomic prediction (GP) utilizing genome-wide molecular marker information to predict breeding values of untested genotypes could help accelerate development of adapted varieties. GP and its implementation through genomic selection is an attractive accompaniment to traditional breeding for its potential to improve selection efficiency, particularly for complex polygenic traits. In this study, we utilized a low-cost ($15/sample) mid-density genetic marker platform developed for blueberries by the Breeding Insight initiative to test the first application of genomic prediction in a breeding population of Northern Highbush Blueberries in the Pacific Northwest (PNW), United States. We predicted breeding values for 18 traits related to fruit quality, timing of flowering, and fruit ripening time using 1,822 markers and phenotype data from 4,320 seedlings. Within-population cross-validation indicated moderate to moderately high prediction accuracy (0.6 to 0.8). In contrast, genome-wide association study (GWAS) identified significant markers with small effect (0.50% < phenotype variance explained (PVE) < 2.5%) for 14 traits, while genomic prediction regularly exceeded 0.60 for most traits in most years. Low GWAS signal and strong genomic prediction performance for several traits are consistent with predominantly polygenic trait architecture in this breeding population, where prediction benefits more from genome-wide relatedness and cumulative small effects than from a small number of detectable major loci. This study demonstrates that related individuals can be predicted when phenotypes from the same environment or year structure help estimate marker effects. These results support the potential for genomic prediction to assist early-stage selection in the US Department of Agriculture-Agricultural Research Service PNW blueberry breeding program, with possible reductions in time and phenotyping burden pending future validation in independent cohorts and breeding cycles.