Unveiling machine learning's impact on in vitro callogenesis optimization in Crocus sativus L

Abstract

Saffron is a spice derived from its flower (Crocus sativus L.) and it is grown for its medicinal and culinary uses worldwide. Callus is an intermediate stage between explant and in vitro organogenesis. The fate of embryogenesis and organogenesis depends on the quality of the callus. Callogenesis is a complex biological process and is affected by various intrinsic and extrinsic factors. Machine learning (ML), can effectively cope with the time and cost constraints of experiments. In this study, we emphasized the effects of sucrose and phytohormones (auxin, and abscisic acid) on callogenesis with the aid of a machine-learning algorithm. Among the different tested hormone combinations, maximum callus proliferation was observed in the MS media plate containing 3 % sucrose, supplemented with 100 µM ABA, 2 and 4 mg/L NAA, and 2,4-D. A total of eight different ML tools were used to predict the efficacy of inputs on callogenesis. Out of all applied algorithms, Xtreme Gradient Boosting (XGB) and Gradient Boosted Regression (GBR) provide better prediction performance than other models on callus induction and proliferation. SHAP analysis reveals the significance of sucrose, plant growth regulators (NAA, 2,4-D, and ABA), and time on callogenesis. Therefore, machine learning programs can effectively save cost and time by guiding experiments and facilitating optimization.