Evaluations of Comet Assay Data Through Statistical Analysis, Machine Learning, and Multi-Criteria Decision-Making Methods for Genotoxic Potential of Food Sweeteners

Although the genotoxic effects of food sweeteners have been studied previously, there is still a lack of application using an integrated approach that combines statistical analysis, Machine Learning (ML), and Multi-Criteria Decision Methods (MCDM) in depth to reveal the DNA-damaging potential of food sweeteners (D-Sorbitol (DS) and Xylitol (XYL)), both alone and in combination (DSX). A dataset on comet assay observations for DNA damage (tail length, tail intensity, and tail moment) was collected from previous studies. Kruskal-Wallis and One-Way ANOVA tests were used to identify significant differences in DNA damage. K-Means and Hierarchical Clustering lead to grouping additives of genotoxic effects, while MOORA and TOPSIS ranked toxicity levels. The findings of MCDM showed that XYL_1000 caused the highest DNA damage (0.726683 and 0.382296). The combination of DS and XYL (DSX_M8) exhibited higher toxicity (0.715258 and 0.37281) compared to their treatments, whereas DSX_M1 revealed the least damaging effect (0.235688 and 0.0324946). This is the first study using this approach. These findings highlight the impact of combining ML and MCDM methods for a more intensive genotoxicity evaluation, providing precious insights into food safety regulations.