利用上皮源性卵巢癌预后的多基因信息建立预测预后模型

Establishment of survival related multigene prognostic model of epithelial ovarian cancer

目的通过基因表达谱汇编(GEO)数据库和癌症基因组图谱(TCGA)数据库筛选并建立与上皮源性卵巢癌(EOC)患者预后有关联的多基因模型,并验证其预后价值。方法从GEO数据库中下载EOC有关芯片数据(GSE14407),筛选出在EOC组织和正常卵巢上皮组织中差异表达的基因(DEGs),采用单因素和多因素Cox回归模型筛选出与预后有关联的DEGs,建立多基因预后模型和预后指数(PI)公式。对TCGA数据库中EOC患者的mRNA数据及临床信息进行整理,通过PI公式对患者进行评分,并根据评分将患者分为低风险组和高风险组。通过Cox回归风险模型分析临床病理参数(年龄、发病位置、组织分级、肿瘤残余及FIGO分期)和预后指数参数与EOC预后的关系。根据年龄、发病位置、组织分级、肿瘤残余及FIGO分期进行分组,采用Kaplan-Meier(K-M)生存分析验证多基因模型对卵巢癌的预后价值。结果共筛选出47个在EOC组织和正常卵巢组织中的DEGs,其中有37个表达下调的DEGs和10个表达上调的DEGs。将上述DEGs进行单因素和多因素Cox回归分析,共筛选出4个DEGs,分别是PACSIN3、KCNT1、LAMP3及KIR3DX1。PI公式:(-0.169×PACSIN3的表达量+0.078×KCNT1的表达量-0.246×LAMP3的表达量-0.147×KIR3DX1的表达量)。Cox回归模型分析证实,年龄、肿瘤残余和预后模型是卵巢癌患者的独立预后因素(P<0.01)。通过K-M生存分析证实,在TCGA数据库的312例EOC患者中,预后评分低风险的患者总体生存期(OS)较高风险患者延长,差异有统计学意义(P<0.05)。在不同的年龄、临床分期、发病位置(单侧和双侧)、肿瘤残余<10 mm的EOC患者亚组中,预后评分低风险的患者OS较高风险患者延长,差异有统计学意义(P<0.05)。结论四基因预后模型是EOC患者的独立预后因素,并在总体和根据各临床病理特征分组的EOC患者亚组中得到了验证。

Objective To establish a multigene prognostic model associated with the prognosis of the epithelial ovarian cancer(EOC) patients and validate the prognostic power through Gene Expression Omnibus(GEO) database and The Cancer Genome Atlas(TCGA) database. Methods The Affymetrix expression profiles(GSE14407) were downloaded and analyzed from GEO dataset. Different expression genes(DEGs) between EOC tissue and normal epithelial tissue were selected to identify a multigene prognostic model. DEGs were selected by univariate and multivariate Cox regression model to establish a multigene prognostic model and prognostic index(PI) formula. mRNA expression data and clinical data of patients were obtained from TCGA dataset. According to PI, the patients were divided into high risk group and low risk group. Clinical and pathological feature(age, neoplasm subdivision, histologic grade, residual tumor, clinical stage) and PI were evaluated as influence factors affecting overall survival(OS) of EOC patients. The prognostic value of multigene model by Kaplan-Meier survival analysis was verified in overall and subgroup EOC patients according to age, neoplasm subdivision, residual tumor, and clinical stage. Results A total of 47 DEGs between EOC and normal tissues were screened, in which 37 DEGs were down-regulated and 17 DEGs were up-regulated. By univariate and multivariate Cox model, 4 DEGs, i.e., PACSIN3,KCNT1,LAMP3 and KIR3 DX1, were finally selected to establish a four-gene prognostic model. PI formula was as follows:(-0.169 × the expression of PACSIN3 + 0.078 × the expression of KCNT1-0.246 × the expression of LAMP3-0.147 × the expression of KIR3 DX1). Age, residual tumor and multigene gene prognostic model were independent prognostic factors of EOC patients(P<0.01). OS of low risk group was longer than high risk group by K-M plots(P<0.05). In different subgroups(age,clinical stage, neoplasm subdivision, residual tumor) of EOC patients, low risk group had longer OS than high risk group(P<0.05). Conclusion Four-gene prognostic model is the independent prognostic factor in EOC patients, and is demonstrated remarkable prognostic value in overall EOC patients and each subgroup according to different clinical and pathological factors.