Taheri-Garavand Amin, Beiranvandi Mojgan, Ahmadi Abdolreza, Nikoloudakis Nikolaos
Mechanical Engineering of Biosystems Department, Lorestan University, Khorramabad, Iran.
Department of Agro-Ecology, Faculty of Agriculture, Lorestan University, Khorramabad, Iran.
BMC Plant Biol. 2025 Jan 6;25(1):19. doi: 10.1186/s12870-024-06044-x.
Savory (Satureja rechingeri L.) is one of Iran's most important medicinal plants, having low irrigation needs, and thus is considered one of the most valuable plants for cultivation in arid and semi-arid regions, especially under drought conditions. The current research was carried out to develop a genetic algorithm-based artificial neural network (ΑΝΝ) model able of simulating the levels of antioxidants in savory when using soil amendments [biochar (BC) and superabsorbent (SA)] under drought. Data under different watering schemes and different levels of soil amendments showed that both BC and SA have mitigating effects over drought stress by optimizing enzymatic and non-enzymatic antioxidant traits (POD, CTA, and APX enzymes). Specifically, using biochar and superabsorbent led to improved homeostasis under water deficit as reflected by lower MDA levels. An ANN model with a 3-10-6 topology was found to be the best model to predict polyphenols (PHE), proline (PRO), peroxidase (POX), catalase (CAT), ascorbate peroxidase (APX) levels, and indicator of oxidative stress malondialdehyde (MDA). The model's efficiency was established using the R-value as the statistical parameter, and simulated GA-ANN data were highly correlated with experimental findings. Across enzymatic antioxidants, APX had the best model fit, having an R-value of 0.9733. On the other hand, POX had a lower predictive correlation (R = 0.8737), indicating a lower capacity of the ANN system in forecasting this parameter. On the other hand, MDA (R = 0.9690) had an elevated assimilation performance over PHE (R = 0.9604) and PRO (R = 0.9245) levels. The current study shows the potential of the ANN model in predicting the content of enzymatic and non-enzymatic antioxidants in savory plants under drought stress as a non-invasive, low-cost experimental alternative.
香薄荷（Satureja rechingeri L.）是伊朗最重要的药用植物之一，灌溉需求低，因此被认为是干旱和半干旱地区最具种植价值的植物之一，尤其是在干旱条件下。当前的研究旨在开发一种基于遗传算法的人工神经网络（ANN）模型，该模型能够模拟在干旱条件下使用土壤改良剂[生物炭（BC）和保水剂（SA）]时香薄荷中抗氧化剂的含量。不同浇水方案和不同土壤改良剂水平的数据表明，BC和SA都通过优化酶促和非酶促抗氧化特性（POD、CTA和APX酶）对干旱胁迫具有缓解作用。具体而言，使用生物炭和保水剂导致水分亏缺下的稳态改善，这表现为较低的丙二醛（MDA）水平。发现具有3-10-6拓扑结构的ANN模型是预测多酚（PHE）、脯氨酸（PRO）、过氧化物酶（POX）、过氧化氢酶（CAT）、抗坏血酸过氧化物酶（APX）水平以及氧化应激指标丙二醛（MDA）的最佳模型。使用R值作为统计参数确定了模型的效率，模拟的GA-ANN数据与实验结果高度相关。在酶促抗氧化剂中，APX的模型拟合最佳，R值为0.9733。另一方面，POX的预测相关性较低（R = 0.8737），表明ANN系统预测该参数的能力较低。另一方面，MDA（R = 0.9690）的同化性能高于PHE（R = 0.9604）和PRO（R = 0.9245）水平。当前研究表明，ANN模型作为一种非侵入性、低成本的实验替代方法，在预测干旱胁迫下香薄荷植物中酶促和非酶促抗氧化剂含量方面具有潜力。
