Poster L11 - Characterizing Brain Tumor Heterogeneity from Omics Analysis in CNS Neoplasms

Introduction: Glioma is the most common type of brain tumor and, despite best efforts, median survival is only 18 months to 2 years. The most aggressive forms: gliosarcoma and glioblastoma, have commonalities in clinical aggressiveness, cellular lineage, and prognostic outlook. The complex landscape of tumor heterogeneity, primarily manifesting within individual tumors, poses a significant challenge to the development and implementation of personalized cancer therapy strategies. Advancements in molecular biology techniques, notably RNA-seq expression analysis, facilitate quantitative measurements of gene expression levels within cells. Leveraging omics data provides a glimpse into the differences in heterogeneity between low-grade gliomas, such as oligodendroglioma and astrocytoma, and high-grade counterparts, including glioblastoma and gliosarcoma.

Materials and

Methods: To study the difference in glioma biology, we used omics data available from the Chinese Glioma Genome Atlas (CGGA). We conducted exploratory data analysis on these datasets to identify genetic changes across tumor grades and gender differences. Identification of significant genes was done by correlation analysis of gene expression data. Supervised Machine Learning classifiers were used to show quantitative differences across tumor groups. K-means clustering algorithm have been used to show class separation across tumor grades.

Results, Conclusions, and Discussions: Glioma biology changes due to various inherent conditions. Classification and clustering algorithms showed low grade gliomas are different from high grade gliomas. Analysis of signaling and metabolic pathways helped identify specific pathways that have most changes in gliomas. Correlation analysis helps us identify genes that are most responsible for glioma development. We also analyzed changes among various gliomas, gender specific differences, and immune interactions based on omics data. Glioma separation classification showed accuracy of 74.4% in average and K-means clustering shows a qualitative separation among various classes. Taken together, this work shows differentiable characteristics among glioma classes indicating future promise of delineation of relationship between gene mutation and pathway changes.

Acknowledgements (Optional):