Exploring Complex Gene Relationships in Brain Cancer Through Graphical Regression

Brain cancer represents a critical health challenge in society, with profound impacts on patients, families, and healthcare systems. Its heterogeneous nature, characterized by diverse genetic profiles and varying tumor behaviors, complicates both diagnosis and treatment. Grasping these complexities is essential for developing effective strategies. This article explores the application of Gaussian graphical models (GGMs) and Graphical Regression (GraphR) to analyze complex dependencies within heterogeneous data, with a particular emphasis on gene expression networks. We introduce an experimental dataset related to brain cancer and describe our methodology for simplifying the data. By employing GraphR, we identify four distinct disease subtypes that contribute to the heterogeneity of data and generate a heatmap to illustrate their impact on the corresponding partial correlations within the network. GraphR facilitates dynamic sparse network estimation, adjusting flexibly to different samplespecific characteristics. Additionally, we use an UpSet plot to visualize intersections and relationships among these factors, further enhancing the accuracy of the gene expression network.