In vitro propagation of Gynura aurantiaca (Blume) DC., a pharmacologically active and anthocyanin-rich ornamental species, was subjected to optimization, using chelated iron treatments and machine learning (ML) modeling. Iron chelates are key modulators of in vitro plant development, influencing morphogenesis, rooting, pigment biosynthesis, and metabolic activity. This study investigated the effects of FeEDTA and graded concentrations of FeEDDHA (100–200–400-800 µM) on micropropagation efficiency, root induction, and pigment accumulation in cultured explants. FeEDDHA treatments significantly influenced the developmental performance of G. aurantiaca explants in a dose-dependent manner. During the micropropagation phase, plant height reached its maximum at 100 µM (9.1 cm) and 800 µM (9.0 cm), while shoot multiplication was most pronounced at 800 µM (6.8 shoots per explant) and 100 µM (5.4). Leaf development peaked at 800 µM with 42.8 leaves per plantlet, compared to 26.8 at 200 µM. Biomass accumulation showed a distinct pattern, with fresh weight highest at 200 µM (4.03 g) and lowest at 400 µM (2.18 g), while dry weight similarly peaked at 200 µM (0.20 g). In the rooting phase, the 200 µM FeEDDHA concentration consistently produced the highest values for plant height (8.7 cm), leaf number (18.2), root length (6.1 cm), and biomass (fresh weight 2.96 g; dry weight 0.17 g). Root number was maximized at 400 µM (39.4), but remained comparably high at 200 µM (35.8). Spectrophotometric pigment analysis further revealed that FeEDDHA treatments significantly enhanced pigment accumulation. The 100 µM concentration notably promoted chlorophyll a, chlorophyll b, and total chlorophyll, while anthocyanin levels peaked at 200 µM in leaf tissues and at 400 µM in stem tissues. These findings indicate that optimal FeEDDHA concentrations are stage- and trait-dependent, with higher doses favoring shoot proliferation, while moderate levels optimize rooting and pigment biosynthesis. In addition to conventional statistical analyses, ML models were employed to predict biological responses based on iron chelate treatments. Random Forest, Support Vector Regression (SVR), and XGBoost algorithms demonstrated high predictive accuracy for pigment concentrations and morphological traits, with R2 values exceeding 0.90 in most models. Feature importance analysis consistently ranked FeEDDHA concentration as the dominant predictor, aligning with experimental trends. This study demonstrates the dual-stage optimization of G. aurantiaca using FeEDDHA and ML-based modeling, offering a scalable framework for enhancing micropropagation protocols and nutrient formulations in plant biotechnology.

Integrating chelated iron treatments and machine learning for optimizing in vitro micropropagation of Gynura aurantiaca (Blume) DC

期刊论文
发表日期：

2026年4月1日

查看全文

Mansur Hakan Erol

摘要

In vitro propagation of Gynura aurantiaca (Blume) DC., a pharmacologically active and anthocyanin-rich ornamental species, was subjected to optimization, using chelated iron treatments and machine learning (ML) modeling. Iron chelates are key modulators of in vitro plant development, influencing morphogenesis, rooting, pigment biosynthesis, and metabolic activity. This study investigated the effects of FeEDTA and graded concentrations of FeEDDHA (100–200–400-800 µM) on micropropagation efficiency, root induction, and pigment accumulation in cultured explants. FeEDDHA treatments significantly influenced the developmental performance of G. aurantiaca explants in a dose-dependent manner. During the micropropagation phase, plant height reached its maximum at 100 µM (9.1 cm) and 800 µM (9.0 cm), while shoot multiplication was most pronounced at 800 µM (6.8 shoots per explant) and 100 µM (5.4). Leaf development peaked at 800 µM with 42.8 leaves per plantlet, compared to 26.8 at 200 µM. Biomass accumulation showed a distinct pattern, with fresh weight highest at 200 µM (4.03 g) and lowest at 400 µM (2.18 g), while dry weight similarly peaked at 200 µM (0.20 g). In the rooting phase, the 200 µM FeEDDHA concentration consistently produced the highest values for plant height (8.7 cm), leaf number (18.2), root length (6.1 cm), and biomass (fresh weight 2.96 g; dry weight 0.17 g). Root number was maximized at 400 µM (39.4), but remained comparably high at 200 µM (35.8). Spectrophotometric pigment analysis further revealed that FeEDDHA treatments significantly enhanced pigment accumulation. The 100 µM concentration notably promoted chlorophyll a, chlorophyll b, and total chlorophyll, while anthocyanin levels peaked at 200 µM in leaf tissues and at 400 µM in stem tissues. These findings indicate that optimal FeEDDHA concentrations are stage- and trait-dependent, with higher doses favoring shoot proliferation, while moderate levels optimize rooting and pigment biosynthesis. In addition to conventional statistical analyses, ML models were employed to predict biological responses based on iron chelate treatments. Random Forest, Support Vector Regression (SVR), and XGBoost algorithms demonstrated high predictive accuracy for pigment concentrations and morphological traits, with R² values exceeding 0.90 in most models. Feature importance analysis consistently ranked FeEDDHA concentration as the dominant predictor, aligning with experimental trends. This study demonstrates the dual-stage optimization of G. aurantiaca using FeEDDHA and ML-based modeling, offering a scalable framework for enhancing micropropagation protocols and nutrient formulations in plant biotechnology.