Application of machine learning models to predict secondary metabolites for the first time in the valuable medicinal plant (Dracocephalum moldavica L.)

Background The aim of this project was to predict, for the first time, the levels of important secondary metabolites during stress in medicinal plants, including carotenoids, amino acid proline, and soluble sugars, without the use of expensive and dangerous chemicals for humans. For this project, artificial neural network models including radial basis function (RBF) and multilayer perceptron (MLP) were used. Results The effectiveness of the models was generally evaluated on experimental datasets through a 5-fold cross-validation method to obtain reliable performance measures. Among the performances were R-squared (R2), mean point error (MAPE), and root mean square error (RMSE). In our study, the RBF model was functionally and optimally better than the MLP model with R2 values ​​of 0.90, 0.975, and 0.941 for soluble sugar, carotenoid pigment, and the highly important amino acid proline, respectively. The associated RMSEs were 0.29, 0.235, and 0.98, while the associated MAPEs were 1.971, 1.124, and 1. 229.The research results demonstrated the RBF model's exceptional ability to effectively represent nonlinear relationships between input variables and biological characteristics. While the MLP model generated plausible forecasts, it was ineffective at estimating the levels of soluble sugars, carotenoids, and proline in the plant. We concluded that the best model was the RBF model, which can be an efficient tool for strategic management of medicinal plants, reducing chemical consumption, reducing environmental pollution, reducing costs, and ultimately developing sustainable agriculture. It can also predict the performance of medicinal plants efficiently and be an effective step towards the sustainable development of agricultural sciences, pharmaceuticals, and the food and medical industries. Conclusions This study emphasizes the value of using computer modeling in agriculture, which is essential for evaluating important aspects of this field, especially with regard to different plant species and environmental conditions.