Gliomas represent 80% of malignant brain tumors. According to histological criteria gliomas are graded from stage I to IV. The median survival time of grade IV patients is 12 to 15 months. It is well-known that molecular characterization based on The Cancer Genome Atlas (TCGA) data allows for concatenating genetic features with prognosis. It is known also from recent pan-glioma analysis for glioma subtypes that clustering of 932 TCGA glioma samples on the basis of DNA methylation has provided more pronounced results than that of 667 TCGA glioma samples profiled using RNA sequencing.

The aim of our study was to establish a highly accurate prediction of patients’ survival using the classification-based feature selection and rule-based modeling. Moreover, we aimed at obtaining specific interdependencies between genes and methylation sites. At the same time, we verified feasibility of the Monte Carlo Feature Selection and Interdependency Discovery (MSFS-ID) algorithm to efficiently process large scale data.