| Popis: |
Background: Breast cancer (BC), the most common cancer among women globally, has been shown by numerous studies to significantly involve non-apoptotic regulatory cell death (RCD) in its pathogenesis and progression. Methods: We obtained the RNA sequences and clinical data of BC patients from The Cancer Genome Atlas (TCGA) database for the training set, while datasets GSE96058, GSE86166, and GSE20685 from The Gene Expression Omnibus (GEO) database were utilized as validation cohorts. Initially, we performed non-negative matrix factorization (NMF) clustering analysis on the BC samples from the TCGA database to discern non-apoptotic RCD-related molecular subtypes. To identify prognostically-relevant non-apoptotic RCD genes (NRGs) and construct a prognostic model, we implemented three machine learning algorithms: lasso regression, random forest, and XGBoost analysis. The expression of selected genes was verified using real-time quantitative polymerase chain reaction (RT-qPCR), single-cell RNA-sequencing (scRNA-seq) analysis, and The Human Protein Atlas (HPA) database. The risk signature was evaluated concerning clinical characteristics and drug sensitivity. Furthermore, we developed a nomogram to predict BC patient survival. Results: The NMF method successfully compartmentalized patients from the TCGA database into three distinct non-apoptotic RCD-related subtypes, with significant variations observed in immune characteristics and prognostic stratification across these subtypes. We identified 5 differentially expressed NRGs used in establishing the risk signature. Patients with different risk groups exhibited distinct clinicopathological features, drug sensitivity, and prognostic outcomes. A nomogram was subsequently developed, incorporating the NRGs-related risk signature, age, T stage, and N stage, to aid clinical decision-making. Conclusion: We identified a novel NRGs-related risk signature, which was expected to become a potential prognostic marker in BC. |
