Jiang Suxiao, Bu Xiangjing, Tang Desheng, Yan Changsheng, Huang Yan, Fang Kun
Department of Surgery, Yinchuan Maternal and Child Health Hospital, Yinchuan, China.
Department of Surgery, The First Affiliated Hospital of Harbin Medical University, Heilongjiang, China.
Front Genet. 2022 Feb 3;12:783026. doi: 10.3389/fgene.2021.783026. eCollection 2021.
Tumor suppressor genes (TSGs) play critical roles in the cell cycle checkpoints and in modulating genomic stability. Here, we aimed to develop a TSG-based prognostic classifier for breast cancer. Gene expression profiles and clinical information of breast cancer were curated from TCGA (discovery set) and Gene Expression Omnibus (GEO) repository (GSE12093 and GSE17705 datasets as testing sets). Univariate cox regression analysis and random forest machine learning method were presented for screening characteristic TSGs. After multivariate cox regression analyses, a TSG-based prognostic classifier was constructed. The predictive efficacy was verified by C-index and receiver operating characteristic (ROC) curves. Meanwhile, the predictive independency was assessed through uni- and multivariate cox regression analyses and stratified analyses. Tumor immune infiltration was estimated ESTIMATE and CIBERSORT algorithms. Small molecule agents were predicted through CMap method. Molecular subtypes were clustered based on the top 100 TSGs with the most variance. A prognostic classifier including nine TSGs was established. High-risk patients were predictive of undesirable prognosis. C-index and ROC curves demonstrated its excellent predictive performance in prognosis. Also, this prognostic classifier was independent of conventional clinicopathological parameters. Low-risk patients exhibited increased infiltration levels of immune cells like T cells CD8. Totally, 48 small molecule compounds were predicted to potentially treat breast cancer. Five TSG-based molecular subtypes were finally constructed, with distinct prognosis and clinicopathological features. Collectively, this study provided a TSG-based prognostic classifier with the potential to predict clinical outcomes and immune infiltration in breast cancer and identified potential small molecule agents against breast cancer.
肿瘤抑制基因（TSGs）在细胞周期检查点以及调节基因组稳定性方面发挥着关键作用。在此，我们旨在开发一种基于肿瘤抑制基因的乳腺癌预后分类器。从TCGA（发现集）以及基因表达综合数据库（GEO）（GSE12093和GSE17705数据集作为测试集）中整理了乳腺癌的基因表达谱和临床信息。采用单变量cox回归分析和随机森林机器学习方法筛选特征性肿瘤抑制基因。经过多变量cox回归分析，构建了基于肿瘤抑制基因的预后分类器。通过C指数和受试者工作特征（ROC）曲线验证了预测效能。同时，通过单变量和多变量cox回归分析以及分层分析评估了预测独立性。使用ESTIMATE和CIBERSORT算法估计肿瘤免疫浸润情况。通过CMap方法预测小分子药物。基于方差最大的前100个肿瘤抑制基因对分子亚型进行聚类。建立了一个包含9个肿瘤抑制基因的预后分类器。高危患者预示着不良预后。C指数和ROC曲线证明了其在预后方面的优异预测性能。此外，这个预后分类器独立于传统的临床病理参数。低危患者表现出免疫细胞如CD8 + T细胞浸润水平增加。总共预测了48种小分子化合物可能用于治疗乳腺癌。最终构建了5种基于肿瘤抑制基因的分子亚型，具有不同的预后和临床病理特征。总体而言，本研究提供了一种基于肿瘤抑制基因的预后分类器，具有预测乳腺癌临床结局和免疫浸润的潜力，并鉴定了针对乳腺癌的潜在小分子药物。
