Recently, the official WeChat account of the Institute of Genomics at the Chinese Academy of Agricultural Sciences announced that the renowned international journal, Nature Genetics, has published the latest research findings by the team led by Professor Zhou Yongfeng from the Shenzhen Agricultural Genomics Research Institute (also the Shenzhen branch of the Lingnan Modern Agricultural Science and Technology Guangdong Provincial Laboratory) of the Chinese Academy of Agricultural Sciences. Leveraging artificial intelligence, this research group has made significant advancements in grape breeding, potentially reducing breeding cycles substantially, enhancing breeding efficiency, and facilitating the precise design and innovation of new grape varieties.
Since 2015, Professor Zhou Yongfeng’s team has focused on the design-based breeding of grapes. In 2023, they successfully published the first complete telomere-to-telomere reference genome map for grapes, with related research featured as a cover article in Horticulture Research. However, a single genome dataset is insufficient for achieving more precise designs. Consequently, the team proceeded to sequence and assemble nine diploid grape varieties, including wild and cultivated species, resulting in 18 telomere-to-telomere haplotype genomes. By integrating existing data, they established Grapepan v1.0, the most comprehensive and accurate grape pan-genome to date, with a total length of 1.43 Gb—nearly three times the size of a single reference genome.
To delve deeper into the relationship between grape genetics and phenotypes, Professor Zhou’s team meticulously selected over 400 representative varieties from nearly ten thousand grape cultivars. Over three consecutive years, they conducted comprehensive investigations into 29 agronomic traits, including cluster size, metabolite content in berries, berry size, and skin color. Based on this data, the team developed grape genotype and phenotype maps and performed quantitative genetic analyses, identifying 148 loci significantly associated with agronomic traits, 122 of which were newly discovered. The study also revealed correlations among loci regulating different traits and significant differentiation regions between various grape populations. These regions contained multiple genetic loci related to agronomic traits, uncovering the genetic basis for the differentiation between wine and table grapes.
To achieve precision breeding, Professor Zhou’s team incorporated machine learning technologies. They developed predictive models to forecast and select early-stage individuals based on scoring, thereby guiding and optimizing breeding strategies. In this study, the team partitioned data encompassing traits and genotypes into three subsets: training, validation, and testing sets. Employing machine learning algorithms to analyze the intricate network relationships between genotypes and phenotypic traits, they constructed the first comprehensive grape genomic selection model. The results demonstrated that integrating structural variation information with machine learning models resulted in a polygenic score prediction accuracy of up to 85%.
The implementation of this model will significantly enhance the efficiency of grape breeding. Breeders can rapidly and accurately assess the genetic potential of vast breeding materials, enabling the prediction of mature traits during the seedling stage. This allows for the early elimination of unsuitable seedlings, thereby reducing unnecessary labor costs and investments. Compared to traditional cross-breeding methods, genomic selection breeding technology can markedly improve grape breeding efficiency, expedite the creation of new grape germplasms, and revolutionize grape breeding strategies.
Currently, the research team has applied for and been granted six national invention patents and has filed one international patent. These research achievements not only represent significant breakthroughs in the field of grape breeding but also provide valuable references and insights for the breeding of other perennial crops.
