Machine Learning- Based Optimization Of Cytotoxicity Testing

This study utilizes machine learning to enhance the cytotoxicity testing of pure zinc. It explores the potential for standardizing such tests across degradable metals and predict the refinement of standard procedures for related in vitro experiments. Data from 51 cytotoxicity experiments on pure Zn are utilized to train and refine five predictive models, i.e., decision tree (DT), random forest, gradient boosted decision tree, support vector machine, and multilayer perceptron. In addition, the impact of biodegradable pure Zn samples on the viability of osteoblasts, endothelial cells, and fibroblasts are assessed. The models are optimized for comparable performance, with the MLP model revealing that osteoblasts had a lower probability of non-toxicity compared to vascular cells across all zinc extract concentrations. Importantly, at concentrations below 40%, all cell types demonstrate a high probability of non-toxicity. “Extract concentration” by the DT model as a critical predictive factor. Cytotoxicity assays indicate that the cell survival rates remain high at Zn extract concentrations up to 40%, beyond which cell activity significantly declines. The results of this study indicate that using extracts of Zn-based alloys at a concentration of 40% enables a reliable assessment of cytotoxicity in accordance with ISO standards. This research offers innovative insights into the cytotoxicity testing protocols for Zn-based biomaterials, elucidating key factors that affect cytotoxicity assessments and defining the limits of in vitro evaluations.