Abstract: Objective Based on TCGA-BRCA data, we integrated cancer stem cell （CSC） related genes with immune-related gene information to construct and evaluate a prognostic risk model for breast cancer patients. Methods Gene expression profiles and clinical data from 1,073 breast cancer patients in The Cancer Genome Atlas （TCGA-BRCA） were collected. Weighted gene co-expression network analysis （WGCNA） was applied to detect co-expression modules of immune-related genes. CSC-associated genes were retrieved from the BCSCdb database and intersected with immune-related genes and differentially expressed genes （DEGs） to obtain 70 candidate genes. Least absolute shrinkage and selection operator （LASSO） Cox regression followed by multivariable Cox regression were used to further select key genes and build a 12-gene prognostic risk model. Model performance was assessed by Kaplan-Meier survival analysis, time-dependent receiver operating characteristic （ROC） curves, and nomogram analysis. Gene set enrichment analysis （GSEA） and immune infiltration analyses were performed to explore underlying biological mechanisms. Results The CSC-related immune prognostic model stratified patients into high- and low-risk groups with significantly different overall survival. Time-dependent ROC analysis indicated good predictive ability for 1-, 3-, and 5-year survival （area under the curve [AUC] = 0.81, 0.77, and 0.75, respectively）. The nomogram demonstrated favorable discriminative performance with a C-index ranging from 0.78 to 0.81. GSEA and immune infiltration analyses provided further insight into pathways and immune features associated with risk stratification. Conclusion We developed a prognostic risk model that integrates CSC-related immune gene expression information in breast cancer. The model effectively discriminated between high- and low-risk patients and shows promising prognostic value.

Key words: breast cancer, cancer stem cells, prognostic model, WGCNA, LASSO regression

摘要： 目的 本研究基于TCGA-BRCA数据,整合乳腺癌肿瘤干细胞相关基因与免疫相关基因信息,构建了一个乳腺癌预后预测风险模型。方法 收集TCGA数据库的1073例乳腺癌患者的基因表达数据与临床资料,应用WGCNA分析鉴定免疫相关基因共表达模块,使用BCSCdb数据库获取肿瘤干细胞（CSCs）相关基因,将其与免疫相关基因和差异表达基因（DEGs）取交集,得到70个候选基因。随后采用LASSO-Cox回归分析和多因素Cox回归分析进一步筛选关键基因,构建了包含12个基因的预后风险模型。通过Kaplan-Meier生存分析、时间依赖性ROC曲线和列线图评估模型预后预测能力,并进一步进行基因富集分析（GSEA）和免疫浸润分析。结果 肿瘤干细胞免疫相关基因预后模型在高低风险组间呈现显著生存差异,ROC曲线分析显示,该模型预测1、3、5年总生存率的曲线下面积（AUC）分别为0.81、0.77和0.75,列线图的C指数为0.78~0.81,显示良好预测性能。结论 本模型整合了乳腺癌CSCs相关免疫基因的信息,可有效区分高低风险患者,具有良好的预后预测价值。

关键词: 乳腺癌, 肿瘤干细胞, 预后模型, WGCNA, LASSO回归