Scholarlink

<p>Sugar beet (<i>Beta vulgaris</i> L.) is a significant source of refined sugar, and its yield heavily depends on successful germination and early seedling establishment. Drought during these stages hampers growth and stand vigor, making genotype-specific evaluation under early water deficit essential for breeding stress-resilient cultivars. In this study, six sugar beet genotypes were evaluated across four polyethylene glycol (PEG) concentrations (0%, 5%, 9%, and 12%) to simulate mild to severe water deficit. Normal seedling count (NS) and early seedling growth traits (root length, shoot length, root and shoot fresh and dry weights) were recorded. Five machine learning algorithms, extreme gradient boosting (XGBoost), random forest (RF), multilayer perceptron (MLP), support vector machine (SVM), and Gaussian process (GP), were applied to model and describe within-experiment relationships between PEG-induced stress intensity, genotype, and early seedling trait variation. PEG-induced stress resulted in significant decreases in all measured traits, with notable genotype-dependent variation. Among the tested algorithms, RF and MLP showed the highest within-dataset modeling performance (<i>R</i><sup>2</sup> ≈ 0.81–0.76), followed closely by XGBoost. In contrast, the kernel-based models GP and SVM achieved moderate performance. Genotype PI590669 exhibited comparatively stronger early seedling performance under severe PEG stress, whereas PI590855 was more sensitive. This study highlights the value of combining physiological traits with machine learning-based modeling to support comparative evaluation of genotype responses under controlled drought conditions. By facilitating a multivariate comparison of genotype responses under PEG-induced drought, this approach provides a framework for the efficient and consistent identification of early-stage stress responses under conditions relevant to the increasing frequency of drought associated with climate change. Future research should extend these methods to multi-environment evaluations, later developmental stages, and integrate genomic data to assess the broader applicability of these findings.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Integrating Machine Learning and Early-Stage Screening to Evaluate Genotype-Specific Seedling Responses to Drought in Sugar Beets (Beta vulgaris L.)

  • 期刊论文

  • 发表日期：

    2026年2月15日

查看全文
  • Omar Gaoua,
  • Mehmet Arslan

摘要

Sugar beet (Beta vulgaris L.) is a significant source of refined sugar, and its yield heavily depends on successful germination and early seedling establishment. Drought during these stages hampers growth and stand vigor, making genotype-specific evaluation under early water deficit essential for breeding stress-resilient cultivars. In this study, six sugar beet genotypes were evaluated across four polyethylene glycol (PEG) concentrations (0%, 5%, 9%, and 12%) to simulate mild to severe water deficit. Normal seedling count (NS) and early seedling growth traits (root length, shoot length, root and shoot fresh and dry weights) were recorded. Five machine learning algorithms, extreme gradient boosting (XGBoost), random forest (RF), multilayer perceptron (MLP), support vector machine (SVM), and Gaussian process (GP), were applied to model and describe within-experiment relationships between PEG-induced stress intensity, genotype, and early seedling trait variation. PEG-induced stress resulted in significant decreases in all measured traits, with notable genotype-dependent variation. Among the tested algorithms, RF and MLP showed the highest within-dataset modeling performance (R2 ≈ 0.81–0.76), followed closely by XGBoost. In contrast, the kernel-based models GP and SVM achieved moderate performance. Genotype PI590669 exhibited comparatively stronger early seedling performance under severe PEG stress, whereas PI590855 was more sensitive. This study highlights the value of combining physiological traits with machine learning-based modeling to support comparative evaluation of genotype responses under controlled drought conditions. By facilitating a multivariate comparison of genotype responses under PEG-induced drought, this approach provides a framework for the efficient and consistent identification of early-stage stress responses under conditions relevant to the increasing frequency of drought associated with climate change. Future research should extend these methods to multi-environment evaluations, later developmental stages, and integrate genomic data to assess the broader applicability of these findings.