Machine Learning Analysis of the Impact of Silver Nitrate and Silver Nanoparticles on Wheat (Triticum aestivum L.): Callus Induction, Plant Regeneration, and DNA Methylation

Abstract

The objective of this study was to comprehend the efficiency of wheat regeneration, callus induction, and DNA methylation through the application of mathematical frameworks and artificial intelligence (AI)-based models. This research aimed to explore the impact of treatments with AgNO3 and Ag-NPs on various parameters. The study specifically concentrated on analyzing RAPD profiles and modeling regeneration parameters. The treatments and molecular findings served as input variables in the modeling process. It included the use of AgNO3 and Ag-NPs at different concentrations (0, 2, 4, 6, and 8 mg L−1). The in vitro and epigenetic characteristics were analyzed using several machine learning (ML) methods, including support vector machine (SVM), random forest (RF), extreme gradient boosting (XGBoost), k-nearest neighbor classifier (KNN), and Gaussian processes classifier (GP) methods. This study’s results revealed that the highest values for callus induction (CI%) and embryogenic callus induction (EC%) occurred at a concentration of 2 mg L−1 of Ag-NPs. Additionally, the regeneration efficiency (RE) parameter reached its peak at a concentration of 8 mg L−1 of AgNO3. Taking an epigenetic approach, AgNO3 at a concentration of 2 mg L−1 demonstrated the highest levels of genomic template stability (GTS), at 79.3%. There was a positive correlation seen between increased levels of AgNO3 and DNA hypermethylation. Conversely, elevated levels of Ag-NPs were associated with DNA hypomethylation. The models were used to estimate the relationships between the input elements, including treatments, concentration, GTS rates, and Msp I and Hpa II polymorphism, and the in vitro output parameters. The findings suggested that the XGBoost model exhibited superior performance scores for callus induction (CI), as evidenced by an R2 score of 51.5%, which explained the variances. Additionally, the RF model explained 71.9% of the total variance and showed superior efficacy in terms of EC%. Furthermore, the GP model, which provided the most robust statistics for RE, yielded an R2 value of 52.5%, signifying its ability to account for a substantial portion of the total variance present in the data. This study exemplifies the application of various machine learning models in the cultivation of mature wheat embryos under the influence of treatments and concentrations involving AgNO3 and Ag-NPs.