Identification of new epithelial cell subpopulations and prognostic biomarkers associated with triple-negative breast cancer-combining machine learning with single-cell analysis

Triple-negative breast cancer (TNBC) is characterized by poor prognosis and elusive etiology. As a major cellular component, epithelial cells play a pivotal role in tumor initiation and progression, yet their heterogeneity and molecular features remain insufficiently defined. In this study, we performed a comprehensive single-cell transcriptomic analysis of TNBC to identify epithelial subpopulations and delineate their molecular and functional characteristics. Using high-dimensional weighted gene co-expression network analysis (hdWGCNA), we constructed co-expression modules and, in combination with multiple machine learning approaches, identified ten key genes. These genes were integrated into a prognostic model, which was rigorously validated by ROC and Kaplan–Meier survival analyses. Comparative profiling of the immune microenvironment and therapeutic responses between high- and low-risk groups revealed significant differences in seven immune cell subsets and eight candidate drugs. Finally, qRT-PCR assays confirmed the differential expression of the ten genes between tumor and normal tissues, thereby reinforcing their clinical relevance.