肺腺癌中预后多维转录组分子标签的构建

Construction of Multi-dimensional Transcriptom Signature for Prognosis of Lung Adenocarcinoma Patients

[目的]通过对TCGA数据库中肺腺癌数据进行挖掘,构建由编码基因(PCG)、长链非编码RNA(lnc RNA)和小RNA(micro RNA)组成的多维转录组分子标签。[方法]采用Cox风险回归、Kaplan-Meier法、随机生存森林、ROC分析等方法,挖掘TCGA癌症公共数据库中肺腺癌转录组二代测序数据,筛选预测效能良好的多维转录组分子标签。[结果]纳入的397例肺腺癌患者的平均年龄为65.67岁,平均生存时间为20.77个月。筛选得到由ELOVL6、RP11-446E9.2、CTD-2555C10.3、PACERR、hsa-mir-140、hsa-mir-31和hsa-mir-582构成的多维转录组分子标签对肺腺癌患者预后预测效能良好。ROC分析其预测效能显示,该分子标签AUC值为0.73,大于TNM分期的0.65(测试组:0.68 vs.0.66)。该分子标签能将肺腺癌患者分成高低风险组,生存时间有显著差异(中位生存时间:25.3个月vs.85.3个月,P<0.001;HR=2.36,95%CI:1.88~2.98,199例)。在测试组Kaplan-Meier分析该多维转录分子组标签也能将患者分成高低风险组(中位生存时间:39.8个月vs.59.3个月,P<0.05,198例)。且多因素Cox回归显示该多维转录组分子标签为独立预后因子。[结论]本研究通过对TCGA数据库的挖掘,构建的多维转录组分子模型对肺腺癌患者预后有良好的指示作用,可作为潜在的肺腺癌患者预后指示标签。

[Purpose] To construct a multi-dimensional transcriptom signature consisting of proteincoding gene（PCG）,long non-coding RNA（lnc RNA）,micro RNA with data-mining the Cancer Genome Atlas（TCGA）public database for prognosis of patients with lung adenocarcinoma.[Methods] Using univariate Cox regression,random survival forest algorithm and ROC analysis,the prognostic markers of lung adenocarcinoma were screened and the multi-dimensional signature was constructed. [Results]The mean age of 397 patients with lung adenocarcinoma was 65.67 years with a mean survival time of 20.77 months.The selected signature was composed by ELOVL6,RP11-446E9.2,CTD-2555C10.3,PACERR,hsa-mir-140,hsa-mir-31,hsa-mir-58,which had highest the area under ROC curve（AUC） in prediction of disease outcome（0.73 Signature vs. 0.65 TNM in the training group and0.68 Signature vs. 0.66 TNM in the test group）. The patients were divided into high-or low-risk group which were significantly associated with survival of lung adenocarcinoma patients in the training group（median survival：25.3 months vs. 85.3 months,P 〈0.001,HR =2.36,95% CI：1.88 ~2.98,n=199）. The signature was applied to the test group,showing similar prognostic values（median survival：39.8 months vs. 59.3 months,P0.05,n=198）. Multivariate Cox regression analysis showed that the signature was an independent prognostic factor for patients with lung adenocarcinoma.[Conclusion] With TCGA data mining,the constructed signature can predict the survival of patients with high accuracy,which may be used as a potential prognostic marker for lung adenocarcinoma.