Machine Learning for Improved Chlorophytum borivilianum Yield: ANNs and GPR in Macronutrient Modelling

Introduction The findings from the in vitro propagation research indicate that the concentration of macronutrients has the most significant impact on shoot organogenesis in plant tissue culture. The present study aims to predict the maximum degree of shoot organogenesis in Chlorophytum borivilianum using two sophisticated computer models: Artificial Neural Network Multi-Layer Perceptron (ANN-MLP) and Gaussian Process Regression (GPR). Methods The data were collected from experiments involving plant cultivation, using 60 explants in a laboratory setting. These experiments included 42 different combinations of macronutrient compositions of Murashige and Skoog (MS) media, and the results related to plant shoot organogenesis were used to train both Artificial Neural Network and Gaussian Process Regression models. The performance of the developed models was evaluated by comparing the observed and predicted output values based on the inputs. Results The results of the output modelling demonstrated that the GPR model exhibits superior accuracy compared to the MLP-ANN model. The model GPR has a percentage accuracy of 99.981 for the number of shoots and 99.885 for the shoot length. On the other hand, the ANN model has an accuracy percentage of 99.825 for the number of shoots and 97.582 for the shoot length. The partial dependence plot further illustrates the relationship between the concentration of macronutrients and the number and length of shoots. Discussion The concentration of macronutrients determines the structural and physiological changes that occur due to interactions between macronutrients and plants. The ANN and GPR models successfully relate the impact of macronutrient concentration on the growth indices. The growth indicators of Chlorophytum borivilianum show a beneficial response to higher doses of calcium chloride and magnesium sulphate. The models show that higher concentrations of potassium nitrate (grams per litre) negatively affect shoot growth, followed by ammonium nitrate. Conclusion The created GPR model can accurately estimate the number of shoots and shoot length by developing various formulations of MS media with variable macronutrient contents for the in vitro propagation of Chlorophytum borivilianum .