基于E2F靶点基因集和免疫亚型差异的肝细胞癌预后风险评分模型的建立

Construction of prognostic assessment model for hepatocellular carcinoma based on E2F targets and immune subtype differences

背景与目的:肝细胞癌(HCC)是肝癌中最常见的种类,HCC患者的预后生存情况较差,其有效的预后预测也面临巨大挑战。许多研究已证实E2F基因家族和免疫微环境相关的基因标志物是癌症的重要预后因素,因此,本研究利用TCGA数据库筛选E2F基因家族和免疫微环境相关的HCC基因标志物,建立新的HCC风险评分模型,并预测HCC潜在治疗靶点。方法:TCGA数据库中下载大型HCC (LIHC)队列(424例样本)。进行了基因集富集分析、基因集单样本富集分析和基因集单样本富集分析分数聚类后的基因表达差异分析,通过Lasso回归筛选标志基因并建模,根据模型计算患者得分并将患者分为高风险组和低风险组。使用受试者工作特征曲线(ROC)、Kaplan-Meier生存曲线、Cox回归分析等多种统计学方法以验证模型的可靠性。所有统计分析均使用R语言软件。最后在Cbioportal数据库查询风险模型的标志基因在TCGA-HCC样本中的基因变异情况,从String数据库中下载蛋白互作信息并用Cytoscape软件进行可视化分析。结果:确认了与HCC密切相关的E2F靶点基因组和免疫相关差异基因后,从中筛选到了与HCC患者总生存率明显相关的7个基因(CYR61,FBLN5,LPA,SAA1,SDC3,SERPINE1,SSRP1),建立7-mRNA预后模型:风险评分=-0.55×CYR61表达-0.18×FBLN5表达-0.17×LPA表达-0.06×SAA1表达+0.31×SDC3表达+0.38×SERPINE1表达+1.08×SSRP1表达。该模型ROC的AUC值为0.846。Kaplan-Meier生存曲线显示,高风险评分患者预后不良(P<0.001),高、低风险评分对预后的区分度与肿瘤大小和UICC分期相似,而比淋巴转移、远处转移和BMI值更好。多因素Cox回归分析显示,7-mRNA预后模型的预测能力独立于临床因素。此外,联合蛋白组学找到7个基因中的关键基因SERPINE1和LPA,预测抑制纤溶酶原激活可能是治疗HCC的新的靶途径。结论:本研究揭示了7个基因与E2F靶点和免疫的相关关系,为HCC患者的不良预后提供了新的生物标志物,并建立了有较高预测准确性预后风险评分模型。然而,多基因预后模型的预测能力仍需大量多中心的循证医学证据证实,被纳入的多基因模型的基因功能和参与的机制仍尚需进行更深入的研究。

Background and Aims: Hepatocellular carcinoma(HCC) is the most common type of liver cancer. The prognosis of HCC patients is poor, and its effective prognosis prediction is also facing significant challenges. Several studies have shown that the genetic markers associated with the E2F gene family and immune microenvironment are important prognostic factors for cancers. Therefore, this study was conducted to screen the HCC gene signatures related to the E2F gene family and immune microenvironment using the TCGA database, establish a new risk assessment model for HCC and predict the potential therapeutic targets for HCC.Methods: A large HCC(LIHC) dataset(n=424) from the TCGA database was downloaded. Gene set enrichment analysis, single sample gene set enrichment analysis, and differential gene expression analysis was performed, marker genes were screened and modeled by Lasso regression, patient scores were calculated according to the model, and patients were divided into high-risk and low-risk groups. Multiple statistical methods, such as the receiver operating characteristic(ROC) curve, Kaplan-Meier survival curve, and Cox regression analysis, were used to verify the model’s reliability. R language software was used for all statistical analyses. Finally, genetic alterations of the marker genes from the risk model were queried in the TCGA-HCC samples in the Cbioportal database. The protein interaction information was downloaded from the String database and visualized in Cytoscape software.Results: After identification of the E2F target genome and immune-related differential genes which were closely related to HCC, seven genes(CYR61, fbln5, LPA, SAA1, SDC3, serpine1, ssrp1) significantly associated with the overall survival rate of HCC patients were screened, and a prognostic 7-mRNA signature model was established: risk score=-0.55×CYR61 expression-0.18×FBLN5 expression-0.17×LPA expression-0.06×SAA1 expression +0.31×SDC3 expression+0.38 ×SERPINE1 expression+1.08×SSRP1 expression The ROC AUC value of the model was 0.846. Kaplan-Meier survival curve showed that patients with high-risk scores had a poor prognosis(P<0.001). The degree of discrimination for prognosis of high and low-risk scores was similar to those of tumor size and UICC stage and higher than those of lymph node metastasis, distant metastasis, and BMI. Multivariate Cox regression analysis showed that the predictive ability of the 7-mRNA signature model was independent of clinical factors. In addition, the key genes SERPINE1 and LPA in the 7 genes were found by combining proteomics, which predicted that inhibiting plasminogen activation was probably a new target approach for treating HCC.Conclusion: This study reveals the correlation between seven genes and E2F targets and immunity, provides new biomarkers for poor prognosis of HCC patients and establishes a prognostic risk score model with high predictive accuracy. However, the predictive ability of the polygenic prognosis model still needs to be confirmed by many evidence-based medical practices from multiple centers, and the gene function and participation mechanism of the included polygenic models still need to be further studied.