Unveiling Key Gene Biomarkers and Clinical Factors in Breast Cancer Subtypes Using Machine Learning and Cox Regression

Objective: Breast cancer is the most frequent malignancy among women, and understanding its genetic subtypes is crucial for improving patient outcomes. Using clinical and gene expression data, this study attempts to uncover important gene biomarkers and clinical factors for various subtypes of breast cancer. Methods: Ten machine learning techniques were initially employed to classify breast cancer subtypes, with the most accurate method serving as the base classifier. Eight feature selection techniques were then used to pinpoint important genes, which were combined with clinical characteristics and subjected to Cox regression analysis to determine their impact on overall survival. Optimal cut-points were identified for each significant gene to stratify data into high and low-expression groups, followed by Kaplan-Meier survival analysis. The selected genes underwent biological validation through Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis, Gene Ontology (GO) enrichment analysis, and a protein-protein interaction (PPI) network. Findings: Age was found to be significantly associated with the hazard as well as survival of all breast cancer subtypes. The genes FOXP1, BRCA1, SETDB1, CHEK2, BRCA1, and MAP2K4 were most strongly associated with the survival time of Basal-Like, HER2, Claudin-Low, Luminal A, Luminal B, and Normal-Like subtypes, respectively. The PPI network identified CDKN2A as the hub protein with the highest node degree. KEGG pathway analysis revealed significant pathways, including the PI3K-Akt, FoxO and ErbB signalling pathways that are closely related to breast cancer. GO analysis identified numerous breast cancer-related biological processes, molecular functions, and cellular components. Novelty: This paper proposes a novel way to find important gene biomarkers and clinical factors by combining advanced feature selection methods with Cox regression. The results' robustness is further increased by applying KEGG Pathway and Gene Ontology analysis for biological validation.