VitisGP links grapevine genomes and phenotypes in a multi-task benchmark

Genomic data for climate adaptation

As climate change reshapes grape-growing conditions, predicting how genomic variation relates to observable vine characteristics is becoming increasingly important. Such predictions could help researchers identify and breed cultivars that may be better suited to emerging environmental pressures.

A new resource called VitisGP approaches this challenge by connecting genome-wide SNP data with grapevine phenotypes. SNPs are variations affecting a single nucleotide in the genome, and machine-learning methods can use these data to investigate relationships between genotype and phenotype.

Twenty-six prediction tasks

VitisGP is presented as a curated, multi-phenotype dataset and benchmark containing 26 heterogeneous phenotype-prediction tasks. These include 11 sommelier-curated aroma traits. The supplied source excerpt also refers to botanical characteristics, but it does not provide the complete task list or detailed definitions of all phenotypes included in the resource.

By bringing multiple target traits into one framework, the benchmark is designed to enable more consistent study across different prediction problems. This contrasts with a common pattern in earlier machine-learning research on grapevine genotype–phenotype relationships, which has often concentrated on individual traits, isolated regions or a limited selection of predictive models.

A shared basis for model evaluation

The multi-task structure could provide a common basis for comparing genotype-to-phenotype modeling approaches and examining whether a method performs consistently across different types of traits. VitisGP’s potential practical relevance lies in supporting cultivar identification and breeding efforts connected to climate adaptation.

However, the material provided does not report detailed model results, accuracy metrics or specific breeding applications. It is therefore not possible to assess comparative performance or practical effectiveness from the available information alone. The original publication was provided by arXiv – Viticulture & Grapevine.