Chinese scientists propose artificial intelligence to promote new methods of grape breeding
Author：Eve Cole
Update Time：2024-11-14 19:30:01
A breeding technology combined with artificial intelligence is significantly shortening traditional grape breeding time. The reporter learned from the Institute of Genomics of the Chinese Academy of Agricultural Sciences that Zhou Yongfeng’s team from the institute proposed a new method of grape breeding using artificial intelligence, which will greatly shorten the breeding cycle, and the prediction accuracy is as high as 85%. Compared with traditional methods, the breeding efficiency can be increased by 400%. . This research is expected to achieve precise grape breeding design, accelerate grape variety innovation, and provide methodological reference for the breeding of other perennial crops. Relevant research results were published in Nature Genetics. Several leaps in grape breeding technology Grape is a favorite fruit of mankind, rich in various nutrients and has various uses. However, it takes three years for a grape seed to germinate and bear fruit. And it takes even longer to develop "satisfactory" grape varieties. Research shows that about 10,000 years ago, people began to try to "transform" grapes. The so-called "transformation" is to selectively improve the original grape traits through specific means. This process is also called breeding.
The history of human domestication of grapes can be traced back to 10,000 years ago. Photo courtesy of the Genome Institute of the Chinese Academy of Agricultural Sciences . In the early days, after people discovered wild grapes, they would preserve high-quality grape seedlings and breed them from generation to generation, leaving only offspring that meet the requirements. This method is useful, but it is highly dependent on Natural germplasm resources can be improved to a limited extent, so they are called Breeding 1.0 technology. Later, people discovered that if they wanted both "high yield" and "high sweetness" of grapes, they could cross the grape varieties with "high yield" and the grape varieties with "high sweetness" to create a hybrid of both parents. Hybrid offspring with excellent traits, this method meets the needs of targeted There is a demand for sexually selected grape varieties, but the breeding cycle is very long, often requiring decades of screening, and the workload is huge. Moreover, because grapes are highly heterozygous, after hybridization, the offspring will have separated traits, and the hybridization effect is not ideal. This approach is called Breeding 2.0 technology. Since the beginning of the 21st century, with the rise of molecular biology, quantitative genetics, bioinformatics and other disciplines, breeders have proposed Breeding 3.0 technology, that is, molecular breeding, which uses molecular markers to "design" traits, and on this basis , proposed Breeding 4.0, that is, intelligent design breeding, which analyzes and predicts based on massive genome and genetic data to improve breeding efficiency and accuracy. Whole-genome selective breeding is the most representative one.
Intelligent design breeding will greatly improve breeding efficiency and accuracy. Photo courtesy of the Genome Institute of the Chinese Academy of Agricultural Sciences. The first grape pan-genome has been released. Currently, grape breeding is still at the 2.0 stage. To achieve the leap from 2.0 to 4.0, we first need sufficiently comprehensive and accurate genomic data. To this end, Zhou Yongfeng's team has focused on the design breeding of grapes since 2015, and will release the first complete telomere-to-telomere reference genome map of grapes in 2023. The relevant research was published as a cover article in Horticulture Research )"superior. However, to achieve precise "design", one genome data is far from enough. On this basis, Zhou Yongfeng's team successively sequenced and assembled 9 diploid grape varieties, including wild and cultivated varieties, and obtained 18 telomere-to-telomere haplotype genomes, and integrated existing Genomic data have been used to construct the first most comprehensive and accurate grape pan-genome, which is nearly three times the size of a single reference genome.
Grape pangenome. Photo provided by the Institute of Genomics, Chinese Academy of Agricultural Sciences. In order to further clarify the relationship between grape genes and traits, Zhou Yongfeng's team selected more than 400 representative grape varieties from nearly 10,000 grape varieties. 29 agronomic traits including metabolite content in berries, berry size and peel color were investigated, and a grape genotype map and trait map were constructed. On this basis, Zhou Yongfeng's team used quantitative genetic analysis to identify 148 loci significantly related to agronomic traits, of which 122 loci were discovered for the first time. The study found that there is a correlation between the loci that regulate different traits, such as the proximity of loci related to soluble solids content and berry width. In addition, there are areas of significant differentiation between different grape groups (wine, table, American table hybrids), with multiple areas related to berry color, skin astringency, berry shape, ear weight, pulp firmness, fruit size, etc. Trait-related genetic loci indicate that divergent selection on agronomic traits promotes the differentiation of winemaking and table grapes.
29 agronomic traits and their correlations among different grape groups. Image provided by the Institute of Genomics, Chinese Academy of Agricultural Sciences. "AI" guides grape breeding. Comprehensive and accurate genomic data is the basis for precise "design" breeding. How to deeply mine these data to optimize breeding strategies and guide breeding? This is a question that must be answered in intelligent breeding. Zhou Yongfeng's team decided to introduce machine learning to build a prediction model to predict and select early individuals based on scores to guide and optimize breeding strategies.
Genomic selection breeding strategies. Photo courtesy of the Institute of Genomics, Chinese Academy of Agricultural Sciences . In this study, the researchers divided the data containing traits and genotypes into three subsets: training set, validation set and test set. Machine learning algorithms were used to analyze the complex network relationships between genotype and trait data, and the first grape genome-wide selection model was constructed using the training data set. The research further adjusted the model parameters through the validation set to optimize the model, and finally the test data set was used to evaluate the final model performance. The results show that the computational polygene score prediction accuracy that combines structural variation information and machine learning models is as high as 85%.
The prediction accuracy of major agronomic traits has been greatly improved. Photo courtesy of the Institute of Genomics, Chinese Academy of Agricultural Sciences. Through this model, breeders can quickly and accurately evaluate the genetic potential of large amounts of breeding materials, thereby better selecting excellent varieties. Compared with cross-breeding, which needs to be judged based on the phenotype of grapes after maturity, whole-genome selective breeding technology can predict the traits of grapes after maturity during the seedling stage, eliminate unqualified seedlings as early as possible, and reduce unnecessary labor costs. and investment, it has great application potential in grape breeding applications, improves grape breeding efficiency, accelerates the creation of new grape germplasm, and innovates grape breeding strategies. At present, relevant research results have been applied for and approved 6 national invention patents and 1 international patent has been applied for. The research was supported by the National Key Research and Development Program, the National Science Fund for Outstanding Youth (Overseas), the National Natural Science Foundation, and the central government’s special funds to guide local science and technology development.
