| dc.description.abstract | As artificial intelligence is steadily rising, the application of machine learning to predict drug responses of cancer cells can be of great value in cancer therapy. Although the drug metformin was originally developed to treat diabetes, its apparent effect on various cancer cells has been widely studied. This study primarily investigates the usage of machine learning algorithms to predict the response of cancer cells to metformin.
A database was constructed from published articles regarding the viability and metabolic changes in cancer cells after exposure to metformin. By applying linear regression to the database, a weak negative linear correlation of – 0.21 was observed between viability and metformin concentration.
Furthermore, various support vector machine algorithms were applied to find models that could make accurate predictions for viability according to the specific variables. From the varying levels of accuracy of the different SVM models, it was evident that the best-suited parameters and kernel functions must be selected to construct accurate models with high performance.
In this study, the colorectal cancer cell line, HCT116, was also exposed to metformin for 24 and 48-hour treatments to directly examine the drug’s influence on cancer cells. To measure viability, alamarBlue assay and CCK-8 assays were conducted on cells treated with various concentrations of metformin. Here, it was apparent that exposure to metformin at higher concentrations for a longer period led to the greatest reduction in cancer cell viability and metabolic activity, displayed by the decreasing trend.
Overall, the usage of machine learning algorithms demonstrated its potential to make highly accurate models that could predict cancer cell's response to metformin treatment. This could further contribute to establishing new cancer treatments. | |
