Machine learning-based optimization of low pressure and putrescine for superior growth parameters and antioxidant metabolites in in vitro Ocimum basilicum seedlings

Low atmospheric pressure is an important abiotic stress in high-altitude and space environments , which can influence plant physiological processes and metabolism. Artificial neural networks (ANNs) are powerful tools to model and predict physiological and biochemical responses of in vitro seedlings. This study is the first to employ ANNs coupled with genetic algorithms to identify the optimal atmospheric pressure , putrescine (Put) level , and sample type to maximize seedling growth and antioxidant metabolites in Ocimum basilicum . The Generalized Regression Neural Network (GRNN) displayed the highest predictive accuracy ( R 2 > 0.953) in both training and testing datasets for all measured physiological responses and antioxidant metabolites compared to other models. In addition , the non-dominated sorting genetic algorithm-II (NSGA-II) was linked to GRNN to access the optimal pressure , putrescine (Put) level , and sample type to obtain the best growth parameters and antioxidant metabolites. Based on the optimization process , a low pressure of 400 mbar with 2.5 mM Put in sprouts is predicted to produce the highest fresh weight (0.46 g) , dry weight (0.02 g) , shoot length (8.81 cm) , root length (9.87 cm) , adventitious root number (38.4) , total phenolic content (8.75 mg GAE g⁻¹ DW) , and anthocyanin content (1552.42 µmol g⁻¹ FW). The validation experiment showed a remarkable rise in physiology and antioxidant metabolites , achieving over a 3.26 , 1.52 , 1.62 , and 2.23-fold rise in fresh weight , dry weight , phenolics , and anthocyanins , respectively. Therefore , GRNN-NSGA-II framework provides an accurate predictive tool for physiological and biochemical parameters and can be a promising candidate multi-factor optimization in medicinal plant tissue culture.