A machine learning-based genomic selection framework for cold-hardy grapevine breeding
Horticulture Research · 13.07.2026
DOI: 10.1093/hr/uhag269 ↗Kurz gefasst
Fachartikel in der Zeitschrift Horticulture Research, online erschienen am 13.07.2026. Verfasst von Xiukun Yao und 12 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Cold hardiness is a critical trait for grapevine cultivation in high-latitude regions, where winter temperatures threaten non-hardy cultivars. Vitis vinifera, the most widely cultivated grape species worldwide, is among the least cold-tolerant members of the Vitis genus and requires burying in order to survive through winter conditions in cold areas such as northern China. As a complex polygenic trait, cold hardiness has few associated genetic markers for marker-assisted selection, hindering breeding progress. Recent advances in genome-wide selection combined with machine learning could provide new opportunities for accurate prediction and selection of this complex trait among the hybrids. Here, we assembled a high-quality phased genome of the cold-hardy V. amurensis cultivar ‘Shuang Hong’. Using resequencing data and evaluating cold hardiness in a core Vitis panel of 227 accessions representing five genetic groups, we performed a genome-wide association study (GWAS) and identified a novel QTL (qCH1) along with other variants associated with cold hardiness. Integration of GWAS and transcriptome meta-analysis pinpointed two candidate genes within qCH1, CIMT1s and HSFA8, as potential negative and positive regulators of cold stress response, respectively. In addition, IBD (Identity By Descent) analysis based on 172 significant variants revealed that several East Asian wild grapes share cold hardiness-associated regions with existing cold-hardy cultivars, highlighting the underexploited breeding potential of East Asian germplasm. Furthermore, machine learning-based genomic prediction achieved 89% accuracy for cold hardiness. Together, these findings provide candidate genes for grapevine cold adaptation and offer a robust predictive framework to support cold-hardy grapevine breeding.
Zitieren
APA
Yao, X., Han, B., Yang, B., Sun, Y., Qiu, J., Liang, J., Liu, J., Xu, Z., Xi, Y., Xu, M., Gao, Q., Lu, J., & Fu, P. (2026). A machine learning-based genomic selection framework for cold-hardy grapevine breeding. Horticulture Research. https://doi.org/10.1093/hr/uhag269
Vancouver
Yao X, Han B, Yang B, Sun Y, Qiu J, Liang J, et al. A machine learning-based genomic selection framework for cold-hardy grapevine breeding. Horticulture Research. 2026. doi:10.1093/hr/uhag269
BibTeX
@article{Yao2026,
author = {Yao, Xiukun and Han, Bin and Yang, Bohan and Sun, Yan and Qiu, Jiayang and Liang, Jianxiang and Liu, Jiaqi and Xu, Zhizhuo and Xi, Yasheng and Xu, Menghao and Gao, Qifei and Lu, Jiang and Fu, Peining},
title = {A machine learning-based genomic selection framework for cold-hardy grapevine breeding},
journal = {Horticulture Research},
year = {2026},
doi = {10.1093/hr/uhag269}
}
EUVIMED