Refinement of Surface Sterilization Protocol for in Vitro Olive (Olea Europaea L.) Shoot Proliferation and Optimizing by Machine Learning Techniques

Abstract

The olive tree (Olea europaea L.) is one of the most ancient fruit species grown throughout history. Given the challenges and costs associated with propagating olive cultivars by cuttings and grafting, it is crucial to identify a method for efficient and widespread propagation. Micropropagation is especially advantageous for propagating plants that are conventionally challenging to propagate or for producing virus-free seedlings or plants with specified traits. This work aimed to improve the in vitro shoot proliferation of O. europaea L. 'Sultani' cultivated in T & uuml;rkiye. Machine learning (ML) techniques were used to predict the efficiency of surface sterilization treatments. The explants were subjected to varied concentrations and durations of five disinfectants: hydrogen peroxide (H2O2), silver nitrate (AgNO3), mercuric chloride (HgCl2), sodium hypochlorite (NaOCl), and chlorine dioxide (ClO2). Each disinfectant was assigned three treatment levels (T1, T2, T3), which varied in concentration and exposure duration. The measured variables were contamination rate, survival rate, growth rate, shoot diameter, shoot length, and leaf number. ClO2 and NaOCl were the most efficient disinfection agents for the growth of explants. ClO2 showed particularly excellent results in terms of shoot diameter (0.765 mm), shoot length (43.733 mm), and leaf number (14.578). NaOCl treatment resulted in the greatest growth percentage (70.55%). AgNO3 treatment performed moderately performance in most parameters, but the lowest contamination rate (13.556%) was observed. Ultimately, the selection of chemical and treatment techniques substantially impacted the efficacy of in vitro olive shoot proliferation. The support vector regression, random forest, extreme gradient boosting (XGBoost), elastic net, and Gaussian processes algorithms were used to model and forecast the optimal sterilizing settings. The XGBoost provided the most accurate (R2) for survival rate, growth rate, shoot diameter, shoot length, and leaf number variables; 0.587, 0.959, 0.843, 0.894, and 0.900, respectively. The XGBoost algorithm was used to predict and optimize surface sterilization. The optimal circumstances for survival and development were projected to include explants sterilized with a 30% solution of NaOCl for 20 min. Moreover, it was projected that explants treated with a 15% concentration of ClO2 for 30 min would be possibly ideal in terms of shoot diameter, shoot length, and leaf number values. ML algorithms could further optimize these protocols for better outcomes, reducing the number of treatments needed and improving efficiency.