Computational Framework For Identifying Hub Genes And Proteins Through Integrative Network Modeling And Machine Learning

Abstract

Breast cancer is a complex disease involving dysregulation of multiple genes, proteins, signaling pathways, and cellular processes. The present study employed an integrative computational systems biology approach to identify differentially expressed genes (DEGs), central hub genes, associated biological pathways, and potential diagnostic biomarkers in breast cancer. Publicly available transcriptomic data comprising breast cancer and normal control samples were analyzed using differential expression analysis, protein–protein interaction (PPI) network construction, network topology, functional enrichment, and receiver operating characteristic (ROC) analysis. Among 19,845 analyzed genes, 756 were significantly upregulated and 706 were significantly downregulated based on |log₂FC| ≥ 1 and adjusted P < 0.05. EGFR, VEGFA, MMP9, STAT3, and CXCL8 were among the most significantly upregulated genes, whereas BRCA1, APC, CDH1, RB1, and PTEN was markedly downregulated. The PPI network comprised 30 protein nodes and 182 interactions, with significant enrichment (P < 1.0 × 10⁻¹⁶). Network analysis identified EGFR as the highest-ranked hub gene, followed by STAT3, PIK3CA, MYC, and TP53. Gene Ontology analysis indicated enrichment in immune response, cell proliferation, angiogenesis, receptor tyrosine kinase signaling, and apoptosis-related processes. KEGG analysis highlighted cancer pathways, PI3K–Akt, MAPK, cytokine–cytokine receptor interaction, focal adhesion, TNF signaling, and apoptosis. ROC analysis demonstrated strong diagnostic performance for EGFR (AUC = 0.912), STAT3 (AUC = 0.886), and PIK3CA (AUC = 0.858). Overall, the integrated computational framework identified biologically important hub genes and pathways that may serve as potential biomarkers and therapeutic targets in breast cancer. Experimental validation and independent patient cohorts are required before clinical application.