By combining near-infrared spectroscopy (NIRS) with machine learning, researchers identified pericarp puncture hardness (PPH) as a reliable physiological marker, leading to a 15% increase in embryo development rate and a 14% boost in plantlet survival.
Seedlessness is an economically valuable trait, with over 85% of cultivated seedless grapes belonging to the pseudo-parthenocarpic type, where embryo growth aborts mid-development. Embryo rescue technology, developed in the 1980s, has been crucial for breeding such varieties by extracting ovules before abortion and cultivating them into seedlings. To date, more than 150 seedless grape cultivars have been developed through this technique. Yet, the greatest challenge remains pinpointing the optimal sampling time for ovules—too early, and embryos are underdeveloped; too late, and they degenerate. Traditional metrics, such as days after pollination (DAP), vary across years and regions, making them unreliable. Accurate, non-destructive methods to determine sampling time have thus become a critical research priority.
A study (DOI: 10.1016/j.plaphe.2025.100044) published in Plant Phenomics on 29 May 2025 by Yan Xu’s & Yingqiu Huo’s team, Northwest A&F University, establishes a non-destructive, NIRS- and AI-based method to determine optimal embryo rescue timing in seedless grapes, significantly improving breeding efficiency and success rates.
In this study, researchers employed a combination of physiological measurements, statistical analysis, and machine learning modeling to precisely determine the optimal sampling time for embryo rescue in seedless grapes. First, they monitored berry and ovule development across six grape cultivars at multiple growth stages, recording parameters such as berry size, fresh weight, soluble solids, color values, and PPH. Principal component analysis (PCA) of 19 indices identified PPH, ovule fresh weight (OFW), and related texture traits as key determinants of embryo degeneration in seedless grapes. To validate their physiological findings, embryo rescue experiments were conducted on ‘Flame Seedless’, ‘Ruby Seedless’, and ‘Jingzaojing’ at six developmental stages, confirming that embryo development rates peaked consistently at the E-L-34 stage. Specifically, the embryo development rate reached 20.34% for ‘Flame Seedless’, 20.34% for ‘Ruby Seedless’, and 16.69% for ‘Jingzaojing’, all markedly higher than rates at earlier or later stages. Correlation and grey relational analyses reinforced PPH as the most reliable non-destructive indicator, with strong positive associations (average r = 0.86, γ = 0.84) with embryo development rates across cultivars. Regression models further defined optimal thresholds: 720 ± 20 g for ‘Flame Seedless’, 990 ± 20 g for ‘Ruby Seedless’, and 633 ± 20 g for ‘Jingzaojing’. Building on these physiological insights, the team processed near-infrared spectral data using seven preprocessing strategies and eight machine learning algorithms, generating 840 models to predict PPH non-invasively. The optimal models—D1+PLSR for ‘Flame Seedless’ and ‘Jingzaojing’, and D1+MLR for ‘Ruby Seedless’—showed high predictive accuracy (R² > 0.79). Applying these models in hybrid crosses demonstrated substantial gains, with embryo development and plantlet rates improving by 8–22% and 8–20%, respectively, compared to traditional methods. Together, these results establish PPH-based, NIRS-assisted modeling as a robust and non-destructive approach for timing embryo rescue, significantly advancing the efficiency of seedless grape breeding.
This research introduces a practical, field-ready tool for grape breeders, offering a reliable, non-destructive way to time embryo rescue. By enhancing embryo and plantlet recovery rates, the method accelerates the breeding cycle of seedless grape cultivars, ultimately benefiting farmers, producers, and consumers worldwide. Beyond grapes, the integration of NIRS and machine learning provides a blueprint for improving embryo rescue and tissue culture techniques across a wide range of fruit crops, where optimal sampling windows are equally critical.
###
References
DOI
10.1016/j.plaphe.2025.100044
Original URL
https://doi.org/10.1016/j.plaphe.2025.100044
Funding information
This work was supported by Shaanxi Province Key Research and Development Plan (2023-YBNY-080), Xi'an Agricultural Technology Research and Development Project (24NYGG0031), the China Agriculture Research System of MOF and MARA (CARS-29-yc-3).
About Plant Phenomics
Plant Phenomics is dedicated to publishing novel research that will advance all aspects of plant phenotyping from the cell to the plant population levels using innovative combinations of sensor systems and data analytics. Plant Phenomics aims also to connect phenomics to other science domains, such as genomics, genetics, physiology, molecular biology, bioinformatics, statistics, mathematics, and computer sciences. Plant Phenomics should thus contribute to advance plant sciences and agriculture/forestry/horticulture by addressing key scientific challenges in the area of plant phenomics.