Identification of cancer-causing gene mutations and gene interactions with artificial intelligence methods

Determination of molecular cancer factors is of critical importance for precision oncology. In this study, first of all, a two-stage mixed-method algorithm was developed using somatic mutation and gene expression data to predict the answers to questions such as which cancer driver genes are, what role they play during the disease, and whether they interact with each other. This hybrid clustering algorithm, based on rule-based control, weighted mutation score, and interquartile range calculations, was tested in 8 different cancer cohorts. The algorithm was compared with 4 different studies published recently. The comparisons showed that this algorithm is better than the 4 studies compared in terms of overall performance. Within the scope of this study, a comparative analysis was conducted on driver genes and gene expressions in central nervous system tumors such as paraganglioma, low-grade glioma, and glioblastoma. As a result of the analyses, the genes ATRX, NF1, MUC16, and TTN were identified as potential driver gene candidates across all three tumor types. FSTL5, GABRG2, VSNL1, and LPL were also identified as common genes with the most significant changes in expression. The findings showed that, these genes were identified as potential therapeutic targets for the treatment of glial and neuroendocrine cell tumors. The final study within the scope of this thesis involves designing a weighted vote-based ensemble classification algorithm for paraganglioma/pheochromocytoma, lowgrade glioma, and glioblastoma tumors using clinical and molecular data. The algorithm was developed using the 6 classifiers with the highest accuracy (Artificial Neural Networks, Logistic Regression, Extra Trees, Random Forest, Gradient Boosting and Extreme Gradient Boosting). As a result of the experiments, an accuracy rate of 90.4% was achieved, indicating strong classification performance. These findings suggest that this method could provide significant support in treatment planning for central nervous system tumors.