Artificial Intelligence and Machine Learning for Predictive Modeling and Hormonal Optimization in Chamomile (Matricaria chamomilla L.) Tissue Culture

Golzari, Rasoul; Maleki-Zanjani, Bahram

doi:10.22034/jmpb.2027.372327.2155

Artificial Intelligence and Machine Learning for Predictive Modeling and Hormonal Optimization in Chamomile (Matricaria chamomilla L.) Tissue Culture

Articles in Press

Document Type : Research Article

Authors

¹ Center of Nobonyad Sirjan, University of Applied Science and Technology, Tehran, Iran

² Department of Agronomy and Plant Breading, Agriculture Faculty, Zanjan University, Zanjan, Iran

10.22034/jmpb.2027.372327.2155

Abstract

Chamomile (Matricaria chamomilla L.) is a high-value medicinal plant with growing pharmaceutical demand, which calls for efficient in vitro propagation systems. This study combined machine learning with classical tissue culture to optimize callogenesis in chamomile explants. Three explant types (cotyledon, root, and node) were cultured on Murashige and Skoog medium supplemented with factorial combinations of indole-3-butyric acid (IBA), 1-naphthaleneacetic acid (NAA), and 6-benzylaminopurine (BAP). The experiment used a completely randomized design with three biological replicates per treatment, testing 18 hormone combinations across the three explant types. Cotyledons showed the highest callogenic potential, with the combination of IBA 1 mg/L + NAA 1 mg/L + BAP 0.1 mg/L producing the maximum callus fresh weight (1.68 g). To further optimize hormone balance, three machine learning models—multilayer perceptron (MLP), generalized regression neural network (GRNN), and random forest (RF)—were tested. MLP achieved the best predictive accuracy (R² = 0.94; RMSE = 0.16), outperforming GRNN and RF. When integrated with a genetic algorithm (MLP–GA), the model identified an optimized hormone combination (0.38 mg/L IBA, 0.29 mg/L NAA, 0.19 mg/L BAP), predicted to yield 1.77 g callus fresh weight, surpassing experimental treatments. The study demonstrates that hybrid ML–GA frameworks are powerful tools for modeling nonlinear plant growth regulator interactions and can be applied as decision-support systems for scalable, reproducible optimization of chamomile tissue culture.

Keywords