基于免疫相关基因构建卵巢癌患者的预后模型

Construction of prognostic model of ovarian cancer patients based on immune-related genes

目的将免疫基因组学与临床特征结合,用R语言等生物信息学方法筛选与卵巢癌预后有关的独立预后因素,并构建卵巢癌预后相关的列线图模型,预测患者生存情况。方法从癌症基因组图谱(The Cancer Genome Atlas,TCGA)数据库获得卵巢癌表达谱数据以及临床数据,标准化处理后通过单因素Cox分析筛选出预后相关基因,并将这些预后相关基因与免疫相关基因数据集取交集,得到与卵巢癌预后相关的免疫基因,通过LASSO回归分析以及多因素Cox回归分析进一步筛选出免疫预后基因构建多因素Cox风险模型。结果经过LASSO回归以及多因素Cox回归分析共筛选出12个预后相关免疫基因,分别为趋化因子CCL配体28(chemokine CCL ligand 28,CCL28)、封闭蛋白4(recombinant claudin 4,CLDN4)、趋化因子CXC配体12(chemokine CXC ligand 12,CXCL12)、内皮细胞特异性分子1(endothelial cell specific molecular-1,ESM1)、鸟苷酸结合蛋白2(guanylate binding protein 2,GBP2)、免疫球蛋白κ可变4-1(immunoglobulin kappa variable 4-1,IGKV4-1)、免疫球蛋白λ可变2-8(immunoglobulin lambda variable 2-8,IGLV2-8)、白细胞介素27受体α(recombinant interleukin 27 receptor alpha,IL27Ra)、peptidase inhibitor 3(PI3)、前列腺素F2α受体(prostaglandin F2αreceptor,PTGFR)、S100钙结合蛋白A5(recombinant S100 calcium binding protein A5,S100A5)和抗原加工相关转运体1(transporter associated with antigen processing 1,TAP1)。基于这些免疫预后基因构建的预测模型的1、2、3和5年的受试者工作特征(receiver operating characteristic,ROC)曲线下的面积(area under the curve,AUC)值分别为0.658、0.748、0.725和0.748。单因素和多因素Cox回归显示,临床分级、年龄以及该模型计算所得的风险评分是卵巢癌患者预后相关的独立因素(均P<0.05)。将临床分级、年龄和风险评分纳入卵巢癌预后相关的列线图模型预测生存期。该模型对卵巢癌患者1、2和3年生存率预测性能的一致性参数(concordance index,C-index)值为0.669。结论基于12个免疫基因计算的风险评分、临床分级以及年龄是卵巢癌患者预后相关的独立因素。本研究构建的免疫基因预后模型可有效预测卵巢癌患者1、2和3年的生存状况。这些免疫基因有望成为卵巢癌免疫治疗的新靶标。

Objective Combined with immunogenomics and clinical features,R language and other bioinformatics methods were used to screen independent prognostic factors of ovarian cancer,and to construct a nomogram model to predict the survival of ovarian cancer patients.Methods The expression profile data and clinical data of ovarian cancer were obtained from The Cancer Genome Atlas(TCGA)database.After standardized processing,prognostic genes of ovarian cancer were screened by univariate Cox analysis,and the intersection of these prognostic genes and immune-related genes were performed to obtain the immune genes associated with the prognosis of ovarian cancer.The immune prognostic genes were further screened out by LASSO regression analysis and multivariate Cox regression analysis to construct a multivariate Cox risk model.Results A total of 12 prognosis-related immune genes were screened out by LASSO regression and multivariate Cox regression,including chemokine CCL ligand 28(CCL28),recombinant claudin 4(CLDN4),chemokine CXC ligand 12(CXCL12),endothelial cell specific molecule-1(ESM1),guanylate binding protein 2(GBP2),immunoglobulin kappa variable 4-1(IGKV4-1),immunoglobulin lambda variable 2-8(IGLV2-8),recombinant interleukin 27 receptor alpha(IL27RA),peptidase inhibitor 3(PI3),prostaglandin F2αreceptor(PTGFR),recombinant S100 calcium binding protein A5(S100A5)and transporter associated with antigen processing 1(TAP1).The area under the receiver operating characteristic(ROC)curve(AUC)values of the 1-,2-,3-and 5-year predictive models based on these immune prognostic genes were 0.658,0.748,0.725 and 0.748,respectively.Univariate and multivariate Cox regression showed that the risk score of the model,clinical grade and age were independent prognostic factors of ovarian cancer(all P<0.05).Clinical grade,age,and risk score were incorporated into a nomogram model related to the prognosis of ovarian cancer to predict survival.The concordance index(C-index)value of the model for predicting the 1-,2-and 3-year survival rates of ovarian cancer patients was 0.669.Conclusions Risk score,clinical grade and age based on 12 immune genes are independent factors related to the prognosis of ovarian cancer.The immune gene prognostic model constructed in this study can effectively predict the 1-,2-and 3-year survival of ovarian cancer patients.These 12 immune genes are expected to be new targets for immunotherapy of ovarian cancer.