Integrated WGCNA and Machine Learning Uncover ECM/Cell Cycle Hub Genes as Prognostic Biomarkers in Lung Adenocarcinoma for Improved Health Care

Introduction: Lung adenocarcinoma (LUAD) is the most common lung cancer subtype, carrying high mortality in global health. Dependable molecular risk assessment tools are urgently needed to complement current clinicopathological staging and improve prognostic accuracy.
Materials and Methods: Using dataset GSE75307, we performed differential expression and weighted gene co-expression network analysis (WGCNA). Functional insights were gained via GO/KEGG over-representation and pre-ranked gene set enrichment analyses. A protein–protein interaction (PPI) network was built using STRING, and Cytoscape/cytoHubba identified hub genes. Consensus machine learning-based feature selection was applied to isolate key prognostic genes, which were then validated using the TCGA-LUAD cohort (n = 515).
Results: We identified 442 unique differentially expressed genes, combining them with tumor-related WGCNA modules to yield 945 candidates enriched in extracellular matrix organization and cell cycle pathways. PPI analysis revealed 17 hub genes, ultimately narrowed down to a robust three-gene signature: TOP2A, CCNB1, and CDK1. In the TCGA-LUAD validation cohort, the model's risk score successfully stratified participant survival risk. Incorporated into a clinicopathological nomogram, it achieved a high concordance index of 0.784, though time-dependent ROC analysis showed modest standalone discrimination. High-risk tumors exhibited an immunosuppressive microenvironment, characterized by elevated M2 macrophage infiltration, upregulation of immune checkpoints (PDCD1, CTLA4), and TGF- β pathway activation.
Conclusion: This study established and validated a dependable three-gene transcriptomic signature (TOP2A, CCNB1, CDK1) that successfully stratifies LUAD survival risk and aligns with immunosuppressive tumor microenvironments, offering a valuable molecular tool to enhance lung cancer prognosis.