Abstract
Lung cancer remains the leading cause of cancer-related deaths worldwide, driven by its complexity and the heterogeneity of its subtypes, which influence pathogenesis, tumor microenvironment, and genetic alterations. We developed a novel weighted gene regulatory network reconstruction method based on maximum entropy and Markov chain entropy principles, which integrates gene expression and DNA methylation data to generate biologically informed networks. Applied to LUAD and LUSC datasets, we define a network methylation index to determine whether gene methylation acts as oncogenic or tumor-suppressive. By revealing a stable core set of pathogenic genes, we identify not only genes with significant expression changes, such as CD74 and HGF, but also pathogenic genes with stable expression, such as BRAF and KDM6A. Additionally, we uncover potential driver genes, such as CORO2B and C20orf194, associated with disease stage, gender, and smoking status. This method offers a more comprehensive understanding of NSCLC mechanisms, paving the way for improved therapeutic strategies.