Artificial Intelligence-based Classification of Glioblastoma Lifespan: Unveiling the Role of TMEM۱۷۶A

Glioblastoma (GBM) is a prevalent and aggressive brain tumor with a grim prognosis,despite extensive research efforts [۱, ۲]. The Cancer Genome Atlas (TCGA) provides a valuabledatabase of gene expression profiles from various cancer types, including GBM, along with clinicaldata and drug response information [۳]. Artificial Intelligence (AI)-based algorithms are extensivelyutilized in biomarker discovery, particularly in analyzing gene expression data.This study utilized data from TCGA, focusing on ۱۵۶ glioblastoma samples. Patients were categorizedinto two groups based on their overall survival (OS): those surviving less or equal to ۱۰۰۰ days (shorterliving batch) and else (longer living batch). The primary objective was to categorize individuals intoshort and long-life groups and identify a biomarker to distinguish between them. Initially, the countmatrix encompassed ۶۰,۴۸۸ genes, and subsequent filtration resulted in ۳۴,۲۲۵ remaining genes persample. Variance Stabilizing Transformation (VST) method using the "DESeq۲" package was thenemployed for data normalization. The Minimum Redundancy Maximum Relevance (MRMR) algorithmselected ۵۰ genes, and a genetic algorithm served as the secondary feature selection method. AIalgorithms, specifically the naïve Bayes algorithm, were employed using the selected features. Astratified ۱۰-fold cross-validation was implemented to ensure robustness, leading to a model accuracyof ۰.۷۰۳ and an AUC score of ۰.۶۷۴. Subsequent analyses, including feature importance assessmentand permutation plots highlighted TMEM۱۷۶A as crucial, showing a slight upregulation in the shortlifegroup compared to the long-life group. To validate its significance, expression patterns werecompared between tumor and healthy samples using cBioPortal, revealing significant overexpressionin tumor samples.In conclusion, this study harnessed AI algorithms to uncover TMEM۱۷۶A as a promising biomarker forthe diagnosis and prognosis of glioblastoma, presenting a potential avenue for improving outcomes inthe management of this formidable brain tumor.

کلیدواژه ها:

Glioblastoma ، Gene expression ، Artificial intelligence ، Biomarker discovery ، TMEM۱۷۶A