联合多数据库构建肝细胞癌自噬基因预后模型

目的构建自噬基因表达特征的肝细胞癌(hepatocellular carcinoma,HCC)患者预后的预测模型。方法从癌症基因组图谱(The Cancer Genome Atlas,TCGA)、基因型-组织表达研究项目(The Genotype-Tissue Expression,GTEx)数据库中分别得到所有HCC和正常肝细胞组织的基因转录表达数据,并将每个样本的基因转录表达数据统一转化为log2(FPKM值+1),消除数据库之间测试平台的数据差异。根据人类自噬基因库中获取的人类自噬基因列表筛选出TCGA-GTEx整合后序列中每个样本对应的自噬基因的表达量。采用R语言limma包以错误发现率(FDR)<0.05及差异倍数|logFC|>1为筛选标准,进行自噬基因差异表达分析。采用R语言clusterProfiler包对差异表达自噬基因以P<0.05为筛选标准,进行基因本体论(Gene Ontology,GO)富集分析及京都基因与基因组百科全书(Kyoto Encyclopedia of Genes and Genomes,KEGG)富集分析。根据自噬基因的差异表达量和患者的临床信息,采用R语言survival包进行单因素的Cox回归分析。进一步将单因素的Cox回归分析中有统计学意义(P<0.05)的自噬基因纳入到多因素Cox回归分析中,以每个差异表达的自噬基因表达量和相对应的回归系数coef值为基础,构建HCC的自噬基因预后模型:expmRNA1×βmRNA1+expmRNA2×βmRNA2+…+expmRNAn×βmRNAn(exp:基因表达量;β:多因素Cox回归分析的回归系数coef)。绘制预测模型的受试者工作特征(receiver operating characteristic,ROC)曲线并计算ROC曲线下面积(area under curve,AUC)评估模型预测价值。结果从TCGA-GTEx数据库中共得到HCC样本374例和正常肝组织样本160例的基因转录表达数据及临床信息。从整合后的样本序列中共筛选出205个自噬基因的表达数据,其中SPNS1、DIRAS3、TMEM74、NRG2、NRG1、IRGM、IKBKE、NKX2-3、BIRC5、CDKN2A、TP73为符合筛选标准的差异表达自噬基因。差异表达自噬基因GO主要富集在丝氨酸/苏氨酸蛋白激酶活性的调控、ErbB 2信号通路、蛋白激酶调节活性、激酶调节活性等功能。差异表达自噬基因主要富集在EGFR酪氨酸激酶抑制剂耐药、Hippo信号通路等KEGG通路。整合并删除生存信息缺失的样本后,共418例样本表达纳入到Cox回归分析中。通过单因素、多因素Cox风险回归分析后,筛选出NRG1(HR=1.5565,95%CI:1.1793~2.0543)、IKBKE(HR=1.7502,95%CI:1.2093~2.5330)这两个自噬基因并建立预后预测模型:(0.44247×NRG1的表达量)+(0.55977×IKBKE的表达量)。预测模型的ROC曲线显示7年总体生存的AUC为0.711。结论基于NRG1和IKBKE表达量构建的HCC预后模型,对HCC患者远期生存率有较高的预测价值。

A prognostic model of autophagy gene in hepatocellular carcinoma based on multidatabase

Objective To construct a prognostic model of hepatocellular carcinoma(HCC)with differential expression of autophagy genes.Method Autophagy genes expression data of HCC and normal liver tissues were obtained from The Cancer Genome Atlas(TCGA)database and The Genotype-Tissue Expression(GTEx)database respectively.The gene expression data from different platforms is normalized into log2(FPKM value+1).Differentially expressed autophagy-related genes of HCC were identified by using R program limma package from the TCGA-GTEx combined data set,the criteria of|logFC|>1 and FDR<0.05 was deemed to be of statistically significance.The Gene Ontology(GO)analyses and Kyoto Encyclopedia of Genes and Genomes(KEGG)pathway analyses were performed by using R program clusterProfiler package,as criteria of P<0.05.Univariate and multivariate Cox proportional hazards regression analyses were performed by using R program survival package to identify the HCC potential prognostic differentially expressed autophagy-related genes.Furthermore,the statistically significant(P<0.05)autophagy genes in the univariate Cox regression analysis were included in the multivariate Cox regression analysis,and the expression of each differentially expressed autophagy gene and the corresponding regression coefficient coef value based on this,the autophagy gene prognosis model of HCC was constructed:expmRNA1×βmRNA1+expmRNA2×βmRNA2+…+expmRNAn×βmRNAn(exp:gene expression level;β:regression coefficient coef of multivariate Cox regression analysis).Draw the receiver operating characteristic(ROC)curve of the predictive model and calculate the area under curve(AUC)to evaluate the predictive value of the model.Results The genes expression data and clinical information of 374 HCC samples and 160 normal liver tissue samples were obtained from TCGA and GTEx databases.Total 205 autophagy genes expression data was obtained from the TCGA-GTEx combined sequence.Among them,SPNS1,DIRAS3,TMEM74,NRG2,NRG1,IRGM,IKBKE,NKX2-3,BIRC5,CDKN2A,TP73 are differentially expressed autophagy genes that meet the screening criteria.GO analysis mainly enriched in"regulation of protein serine/threonine kinase activity","ErbB 2 signaling pathway","protein kinase regulator activity"and"kinase regulator activity";KEGG analysis enriched frequently in"EGFR tyrosine kinase inhibitor resistance","Hippo signaling pathway".After integrating and deleting samples with missing survival information,a total of 418 sample expressions were included in the Cox regression analysis.After univariate and multivariate Cox risk regression analysis,the two autophagy genes NRG1(HR=1.5565,95%CI:1.1793-2.0543)and IKBKE(HR=1.7502,95%CI:1.2093-2.5330)were screened out and a prognostic prediction model was established:(0.44247×NRG1 expression level)+(0.55977×IKBKE expression level).The ROC of the prognosis model shows that the AUC of the overall seven-year survival is 0.711.Conclusion The prognosis model of HCC based on NRG1 and IKBKE has high predictive value for the long-term survival rate of hepatocellular carcinoma patients.