Leveraging machine learning to unravel the impact of cadmium stress on goji berry micropropagation

• Published in: PloS one 2024-06, Vol.19 (6), p.e0305111
• Main Authors: Isak, Musab A, Bozkurt, Taner, Tütüncü, Mehmet, Dönmez, Dicle, İzgü, Tolga, Şimşek, Özhan
• Format: Article
• Language: English
• Subjects: Agricultural practices; Agricultural production; Agricultural research; Algorithms; Antioxidants; Artificial intelligence; Berries; Biology and Life Sciences; Cadmium; Cadmium - toxicity; Cadmium content; Computer and Information Sciences; Efficiency; Food; Food safety; Fruit - drug effects; Fruit - growth & development; Gaussian process; Genotype; Genotypes; Heavy metals; Learning algorithms; Lycium - drug effects; Lycium - growth & development; Lycium barbarum; Machine Learning; Medicine and Health Sciences; Micropropagation; Multilayer perceptrons; Neural networks; Physical Sciences; Physiological aspects; Physiology; Plant growth; Plant Roots - drug effects; Plant Roots - growth & development; Plant tissues; Prediction models; Research and Analysis Methods; Safety and security measures; Statistical methods; Stress, Physiological - drug effects; Support vector machines; Sustainable agriculture; Sustainable practices; Tissue culture; Toxicity
• ISSN: 1932-6203; 1932-6203
• DOI: 10.1371/journal.pone.0305111

This study investigates the influence of cadmium (Cd) stress on the micropropagation of Goji Berry (Lycium barbarum L.) across three distinct genotypes (ERU, NQ1, NQ7), employing an array of machine learning (ML) algorithms, including Multilayer Perceptron (MLP), Support Vector Machines (SVM), Random Forest (RF), Gaussian Process (GP), and Extreme Gradient Boosting (XGBoost). The primary motivation is to elucidate genotype-specific responses to Cd stress, which poses significant challenges to agricultural productivity and food safety due to its toxicity. By analyzing the impacts of varying Cd concentrations on plant growth parameters such as proliferation, shoot and root lengths, and root numbers, we aim to develop predictive models that can optimize plant growth under adverse conditions. The ML models revealed complex relationships between Cd exposure and plant physiological changes, with MLP and RF models showing remarkable prediction accuracy (R2 values up to 0.98). Our findings contribute to understanding plant responses to heavy metal stress and offer practical applications in mitigating such stress in plants, demonstrating the potential of ML approaches in advancing plant tissue culture research and sustainable agricultural practices.