BACKGROUND: In vitro callus culture systems have been well documented to increase the production of bioactive metabolites. However, optimization of culture conditions through traditional methods consumes time and resources. In the current study, we utilized Machine Learning (ML) tools to improve callus formation conventions. The resultant callus and the leaf tissues of the mother plant were analyzed for their phytochemical profiles to gain insights into the altered synthesis of metabolites. METHODS: Leaf explants were grown in the dark on Murashige & Skoog medium (MS) with 28 altered plant growth regulator combinations. The parameters of callogenesis were measured and analyzed using a generalized regression neural network (GRNN) to predict optimal hormone concentrations. Predicted concentrations were validated through in vitro experiments. Metabolite quantification assays and profiling were carried out through liquid chromatography-mass spectrometry (LC-MS). RESULTS: The mean earliest callus initiation was calculated as 14.9 days in MS media fortified with 3 mg/L 2,4-D and 0.5 mg/L BAP. The maximum callus mean fresh weight was recorded as 2.948 g in MS media with 2.5 mg/L 2,4-D and 1.0 mg/L BAP. The data obtained through experimentation were fed into a machine learning model to predict the optimal concentrations for callus initiation and maximum callus fresh weight. Machine learning predicted the earliest callus initiation as the 14th day if grown in MS media with 2.92 mg/L 2,4-D, 0.35 mg/L Kin, and 0.17 mg/L BAP, which closely aligns with validated experimental results showing 15.1 days. The predicted callus fresh weight of 2.954 g in MS media with 2.5 mg/L 2,4-D, 0.11 mg/L Kin, and 0.91 mg/L BAP on validation was in sync with the experimental results, amounting to 3.004 ± 0.098 g, and hence was used for subculturing. Phytochemical analysis indicated notable increases in phenolic (1.22-fold) and flavonoid (1.27-fold) contents in callus extracts compared to mother plant leaves, whereas terpenoid levels were lower. CONCLUSION: The current study demonstrates the effective incorporation of ML in optimal callogenesis. It highlights the improved accumulation of phenolics and flavonoids in calli obtained from ML optimized conditions, emphasizing its precise prediction in maximizing output. These findings can contribute to the development of efficient biotechnological strategies for the production of pharmaceutically important phytochemicals.
Purohit et al. (Thu,) studied this question.