Classification of glioma subtypes by mRNA expression data using machine learning methods

Gliomas are one of the most common types of brain tumors. Histopathologically, they are divided into two groups as low-grade and high-grade gliomas. Glioblastoma multiforme (GBM), a malignant type with rapid growth and early spread, is an extremely aggressive tumor with high mortality and recurrence rates. Although the exact cause of gliomas is unknown, despite different treatment approaches, poor prognosis and recurrence are generally inevitable. Therefore, correct classification, early diagnosis and effective treatment of gliomas are of great importance. In order to make personalized treatment planning, the tumor grade must be determined correctly. In this thesis study, it is aimed to classify tumors as high-grade (Glioblastoma Multiforme-GBM) and low-grade gliomas (Low-Grade Glioma-LGG) by using mRNA expression data of patients diagnosed with glioma. Machine learning is an effective method that allows the analysis of genetic data at the molecular level and is widely used for big data analysis. In this context, Naive Bayes and support vector machine, which are machine learning methods, were used in the thesis study. The most successful prediction model developed was determined as radial-based support vector machine with 100% sensitivity, 98% specificity and 99% accuracy values. It will contribute to personalized treatment approaches thanks to the correct classification of gliomas according to their subtypes. Keywords: Classification, Genetics, Glioma, Machine learning, mRNA, Support vector machine