Prediction and validation of <i>in vitro</i> callogenesis of fenugreek (<i>Trigonella foenum-graecum</i> L.) plant with different machine learning algorithms

İncetekin M., Uranbey S.

EURO-MEDITERRANEAN JOURNAL FOR ENVIRONMENTAL INTEGRATION, cilt.11, sa.6, 2026 (ESCI, Scopus)

Özet

In this study, different explant types (cotyledon, petiole, and root), two different fenugreek varieties (& Ccedil;ift & ccedil;i and G & uuml;rarslan), and different concentrations of 2,4-D (0, 1.13, 2.26, 4.52, 9.05, 18.10 & micro;M) were used. The effects of these parameters on callus formation and callus weight were investigated. Callus formation was not observed in the nutrient medium without 2,4-D and in the nutrient medium containing 18.10 & micro;M 2,4-D. Good results were obtained for both the callus formation parameter and the callus weight parameter from nutrient media containing 1.13 and 2.26 & micro;M 2,4-D. The study resulted in the creation of a dataset containing 144 data points. The explant type, varieties and 2,4-D concentrations were used as input variables, while callus weight and callus formation rates were used as output variables. The obtained data were estimated and validated using different machine learning methods (k nearest neighbours (k-NN), random forest (RF), multilayer perceptron (MLP), support vector regression (SVR), extreme gradient boost (XGBoost)). Cross-validation was performed using the k-fold method, and a k-fold value of 5 was chosen for all models. For the training and test sets, this technique divided the dataset into ten equal segments at a ratio of 4:1. R2, RMSE, MAE, MAPE validation metrics were used to evaluate and compare the models used. The XGBoost model demonstrated the highest performance for callus formation with an R & sup2; value of 0.9115. Additionally, the machine learning method XGBoost was determined to be the one with the highest performance within the callus weight parameters.