Scholarlink

<p>This study investigated the effects of gibberellic acid (GA₃) treatments and salinity stress on germination and seedling development in <i>Festulolium</i> cultivars. Seeds were exposed to salinity levels of 0, 5, 10, 15, and 20 dS/m and treated with distilled water, 100 ppm, or 200 ppm GA₃ for 12&#xa0;h. Germination percentage (GR), mean germination time (MGT), shoot height (SH), root length (RL), seedling fresh weight (FW), dry weight (DW), seedling vigor index (SVI), and proline content were assessed. Salinity stress markedly reduced GR, SH, RL, and SVI, while increasing MGT and proline. Severe reductions in germination and seedling growth were observed at 15–20 dS/m. At 20 dS/m, SVI was higher in Lofa (9.63) than in Hostyn (7.96). GA₃ at 200 ppm improved GR, SH, and SVI at 0–10 dS/m and enhanced proline accumulation under high salinity. To complement experimental findings, four machine learning algorithms—k-Nearest Neighbors (kNN), Multilayer Perceptron (MLP), Random Forest (RF), and XGBoost—were evaluated for predictive modeling of seven traits. The MLP model outperformed others, achieving the highest accuracy (R² = 0.66–0.97; NRMSE = 0.04–0.14), with near-perfect predictions for SH and RL (R² = 0.97; CCC = 0.98–0.99). kNN performed well for SH, RL, and MGT (R² = 0.83–0.90) but poorly for GR and proline. RF and XGBoost showed moderate accuracy (R² ≈ 0.85–0.92 for morphological traits) but weak predictability for biochemical parameters (R² = 0.29–0.54). Overall, morphological traits were predicted more accurately than biochemical ones, highlighting the superior generalization ability of MLP in capturing nonlinear responses to salinity and GA₃ treatments. In conclusion, 200 ppm GA₃ improved seedling performance under salt stress, and machine learning, particularly MLP, proved to be a powerful tool for predicting cultivar responses, offering a reliable framework for stress physiology studies in <i>Festulolium</i>.</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

Integrative Machine Learning Assessment of Gibberellic Acid Effects on Germination and Seedling Growth of Festulolium Cultivars Under Salt Stress

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月3日

查看全文
  • Hilda Farida,
  • Satı Uzun,
  • Onur Okumus,
  • Akife Dalda Şekerci,
  • Musab A. Isak,
  • Özhan Şimşek

摘要

This study investigated the effects of gibberellic acid (GA₃) treatments and salinity stress on germination and seedling development in Festulolium cultivars. Seeds were exposed to salinity levels of 0, 5, 10, 15, and 20 dS/m and treated with distilled water, 100 ppm, or 200 ppm GA₃ for 12 h. Germination percentage (GR), mean germination time (MGT), shoot height (SH), root length (RL), seedling fresh weight (FW), dry weight (DW), seedling vigor index (SVI), and proline content were assessed. Salinity stress markedly reduced GR, SH, RL, and SVI, while increasing MGT and proline. Severe reductions in germination and seedling growth were observed at 15–20 dS/m. At 20 dS/m, SVI was higher in Lofa (9.63) than in Hostyn (7.96). GA₃ at 200 ppm improved GR, SH, and SVI at 0–10 dS/m and enhanced proline accumulation under high salinity. To complement experimental findings, four machine learning algorithms—k-Nearest Neighbors (kNN), Multilayer Perceptron (MLP), Random Forest (RF), and XGBoost—were evaluated for predictive modeling of seven traits. The MLP model outperformed others, achieving the highest accuracy (R² = 0.66–0.97; NRMSE = 0.04–0.14), with near-perfect predictions for SH and RL (R² = 0.97; CCC = 0.98–0.99). kNN performed well for SH, RL, and MGT (R² = 0.83–0.90) but poorly for GR and proline. RF and XGBoost showed moderate accuracy (R² ≈ 0.85–0.92 for morphological traits) but weak predictability for biochemical parameters (R² = 0.29–0.54). Overall, morphological traits were predicted more accurately than biochemical ones, highlighting the superior generalization ability of MLP in capturing nonlinear responses to salinity and GA₃ treatments. In conclusion, 200 ppm GA₃ improved seedling performance under salt stress, and machine learning, particularly MLP, proved to be a powerful tool for predicting cultivar responses, offering a reliable framework for stress physiology studies in Festulolium.