Genomic prediction of phenological and fruit-quality traits in a multi-family Japanese plum breeding population

The genetic improvement of the Japanese plum has been mainly hampered by the long juvenile periods of trees, driving the search for alternatives to increase the efficiency in developing new cultivars. However, the high degree of synteny among Prunus spp. has facilitated the identification of genetic markers associated with the variation of desirable traits. In this context, genomic selection (GS) has emerged as a promising approach for predicting phenotypic traits of relevance in Prunus spp. This study employed the GS approach to predict five traits in a multiple-family Japanese plum population, evaluated across two agricultural seasons. For this, 1062 trees were genotyped (11K SNPs) and 963 phenotyped for the following traits: flowering beginning date (FBD), harvest date (HD), fruit weight (FW), soluble solids content (SSC), and acidity (Ac). Six traditional parametric and non-parametric GS methods were used to estimate the variance components of phenotypic traits and assess their predictive ability (PA). Additionally, a dimensionality reduction model of the predictor variables was evaluated. According to the models with the best goodness-of-fit, the genomic heritability estimates ranged from 0.34 (Ac) to 0.75 (HD). No significant differences in PA values were observed across methods. Except for Ac and SSC, most studied traits were predicted with a PA value > 0.6 (FW, FBD, HD). The dimensionality reduction procedure significantly increased the PA for all traits. For Ac, PA was up to twice as high when using the variable selection approach, while for SSC, the PA increased by up to 50% using 14% of the total SNPs. This study demonstrated the potential of GS for predicting phenological and fruit quality traits in Japanese plum, offering a new strategy to enhance the efficiency of fruit tree breeding and a brighter future for improving fruit quality in new varieties developed through this approach.