Silver nanoparticles (Ag-NPs) demonstrate unique properties and their use is exponentially increasing in various applications. The potential impact of Ag-NPs on human health is debatable in terms of toxicity. The present study deals with MTT(3-(4, 5-dimethylthiazol-2-yl)-2, 5-diphenyl-tetrazolium-bromide) assay on Ag-NPs. We measured the cell activity resulting from molecules' mitochondrial cleavage through a spectrophotometer. The machine learning models Decision Tree (DT) and Random Forest (RF) were utilized to comprehend the relationship between the physical parameters of NPs and their cytotoxicity. The input features used for the machine learning were reducing agent, types of cell lines, exposure time, particle size, hydrodynamic diameter, zeta potential, wavelength, concentration, and cell viability. These parameters were extracted from the literature, segregated, and developed into a dataset in terms of cell viability and concentration of NPs. DT helped in classifying the parameters by applying threshold conditions. The same conditions were applied to RF to extort the predictions. K-means clustering was used on the dataset for comparison. The performance of the models was evaluated through regression metrics, viz. root mean square error (RMSE) and R2. The obtained high value of R2 and low value of RMSE denote an accurate prediction that could best fit the dataset. DT performed better than RF in predicting the toxicity parameter. We suggest using algorithms for optimizing and designing the synthesis of Ag-NPs in extended applications such as drug delivery and cancer treatments.
Keywords: Ag-NPs; Decision Tree; HEK293; PC 12 cell line; Random Forest; cytotoxicity; k-means clustering; machine learning; regression metrics.