引用本文
  • 罗 安,朱欣彦,胡晔东,等.基于肿瘤基质评分的胃癌预后基因分析[J].同济大学学报(医学版),2020,41(4):418-425.    [点击复制]
  • LUO An,ZHU Xin-yan,HU Ye-dong,et al.Analysis of prognosis-related genes of gastric cancer based on tumor stromal score[J].同济大学学报(医学版),2020,41(4):418-425.   [点击复制]
【打印本页】 【在线阅读全文】【下载PDF全文】 查看/发表评论下载PDF阅读器关闭

←前一篇|后一篇→

过刊浏览    高级检索

本文已被:浏览 30次   下载 15 本文二维码信息
码上扫一扫!
基于肿瘤基质评分的胃癌预后基因分析
罗安,朱欣彦,胡晔东,刘雁冰,冉晨曦,刘菲
0
(同济大学附属东方医院消化内科,上海 200120)
摘要:
目的 分析比较不同肿瘤基质评分胃癌患者的基因表达特征,鉴定与评分相关的胃癌预后基因,以期为临床胃癌诊断和预后提供更精准的手段。方法 从癌症基因组图谱数据库(the cancer genome atals, TCGA)下载胃癌的临床资料和组织转录组测序(ribonucleic acid sequencing, RNAseq)表达数据。从基质免疫评估数据库(estimation of stromal and immune cells in malignant tumor tissues using expression data, ESTIMATE)网站下载TCGA数据库中胃癌患者基质评分信息。获取患者的临床信息、RNAseq表达谱、基质评分。按照基质评分的高低分为高基质评分组和低基质评分组,分析基质评分与胃癌预后的关系。用R语言DEseq2包进行标准化处理和差异分析;WGCNA(weight correlation network analysis, WGCNA)包筛选与基质评分密切相关的差异基因;单因素COX风险比例回归模型(COX proportional model, COX)初步筛选基质评分密切相关基因中与胃癌预后相关的基因;LASSO(least absolute shrinkage and selection operator, LASSO)回归模型筛选出其中影响胃癌预后的关键基因,计算最小λ值;多因素COX回归分析构建关键基因胃癌预后模型,并量化基因表达量与患者生存时间的关系;模型内部绘制关键基因的生存曲线。最后通过其他公共数据库(KM-plotter数据库和Oncomine数据库)验证这些基因在胃癌大样本的表达和预后。结果 基质评分越高的患者表现为预后更差(P<0.05)。对患者的RNA-seq差异表达分析筛选得到1581个差异表达基因;从中通过WGCNA筛选出1015个基因与胃癌基质评分密切相关;单因素COX回归选出377个基因与胃癌患者预后相关(P<0.05);LASSO回归筛选出12个与胃癌预后相关的关键基因,最小λ=12;多因素COX回归分析显示该模型C指数为0.68,3年生存期和5年生存期的预测值基本贴合实际值,3年生存期曲线下面积(area under the curve, AUC)为0.693,5年生存时间AUC为0.725。12个基因中,ACTA1、ADAMTS12、LINCO1614、MATN3、MTUS2、PLCL1、POSTN、SERPINE1、TPTEP1表达量越高,患者生存期越短,GAD1和MMP16表达量的越低,患者生存期越短;6个基因(ADAMTS12、MATN3、MEGF10、PLCL1、POSTN、SERPINE)各自作为独立危险因素,具有最佳的胃癌预后预测功能(P<0.05)。KM-plotter数据库和Oncomine数据库符合本研究的预测结果。结论 肿瘤基质评分越高的胃癌患者,预后更差、生存周期更短。6个基因ADAMTS12、MEGF10、PLCL1、POSTN、MATN3、SERPINE与患者的肿瘤基质评分及预后密切相关。其表达越高,患者评分越高,预后越差、生存周期越短。本研究鉴定了与胃癌基质评分相匹配的预后基因,提示胃癌基质研究的进一步方向。
关键词:  胃癌  高通量测序  癌症基因组图谱数据库  权重共表达分析  预后基因
DOI:10.16118/j.1008-0392.2020.04.003
投稿时间:2020-04-25
基金项目:国家自然科学基金(81970358)
Analysis of prognosis-related genes of gastric cancer based on tumor stromal score
LUO An,ZHU Xin-yan,HU Ye-dong,LIU Yan-bing,RAN Chen-xi,LIU Fei
(Dept. of Gastroenterology, East Hospital, Tongji University School of Medicine, Shanghai 200120, China)
Abstract:
Objective To identify the prognosis-related genes of gastric cancer based on bioinformatics analysis. Methods Clinical and tissue transcriptome sequencing(RNAseq)data of gastric cancer were downloaded from the cancer genome atlas(TCGA). The gastric cancer stromal scores of patients was downloaded from the database of estimation of stromal and immune cells in malignant tumor tissues using expression data(ESTIMAT). The clinical information, RNAseq expression profile and tumor stromal score of the patients were integrated. The relationship between stromal score and the prognosis of gastric cancer was analyzed. The DEseq2 package was used for standardization and differential expression analysis of the RNAseq matrix data; the WGCNA(weight correlation network analysis)package was used to select the gene module which was closely related to stromal score among the differential expression genes. Uunivariate COX regression model was used to screen out the genes closely associated with prognosis of gastric cancer; Lasso(least absolute shrinkage and selection operator) regression model was used for screening out the key genes influencing the prognosis of gastric cancer, and the minimum λ of the model was calculated. The multivariate COX regression analysis was performed to construct the prognostic model of key genes in gastric cancer and to analyze the relationship between gene expression and survival time of patients. The survival curves of key genes were generated. Finally, the expression and prognosis of these genes were verified in large samples of gastric cancer from other public databases(km-plotter database and Oncomine database). Results The prognosis of gastric cancer patients with high stromal score were significantly worse that of that of patients with low score(P<0.05). Based on RNAseq data between gastric cancer patients with high and low stromal scores, 1581 differentially expressed genes were screened out. WGCNA picked out a gene module containing 1015 genes which were closely related to the stromal score of gastric cancer. The univariate COX regression showed that 377 genes in the gene module were associated with the prognosis of gastric cancer patients(P<0.05); 12 key genes related to the prognosis of gastric cancer under the analysis of the Lasso regression, with λ=12. The multivariable COX regression analysis showed that C index of the model was 0.68, the predicted values of 3-year survival and 5-year survival were roughly consistent with the actual survival of patients. The area under the ROC curve(AUC) was 0.693 for 3 year survival, AUC was 0.725 for 5 year survival, respectively. Among the 12 key genes, ACAT1, ADAMTS12, LINCO1614, MATN3, MTUS2, PLCL1, POSTN, SERPINE1, TPTEP1 were positively correlated with the survival of patients, while GAD1, MMP16 had negative correlation. Six genes(ADAMTS12, MATN3, MEGF10, PLCL1, POSTN, SERPINE) had the best prognostic function of gastric cancer(P<0.05), when acting as an independent risk factor. The result was confirmed by Km-plottor and Oncomine database. Conclusion The stromal score is negatively correlated with the prognosis and the survival time of gastric cancer patients; meanwhile the expressions of ADAMTS12, MEGF10, PLCL1, POSTN, MATN3 and SERPINE genes are positively related to the stromal score and the prognosis of patients. The study provides information for further stromal research in gastric cancer.
Key words:  gastric cancer  RNAseq  the cancer genome atlas  weight correlation network analysis  genes of predict prognosis

您是第2184913位访问者
版权所有《同济大学学报(医学版)》编辑部
主管单位:教育部 主办单位:同济大学
地  址: 上海四平路1239号 邮编:200092 电话:021-65980705 E-mail: yxxb@tongji.edu.cn
本系统由北京勤云科技发展有限公司设计