依据体内基因表达数据构建子宫颈癌辅助放疗敏感模型初探

Preliminary investigation into construction of sensitivity model of adjuvant radiotherapy in cervical cancer based on gene expression data in vivo

目的基于子宫颈癌组织基因表达情况构建放疗敏感模型,探讨应用模型判断辅助放疗获益人群的可行性。方法利用癌症和肿瘤基因图谱(TCGA)子宫颈癌数据库选取行单纯根治性手术切除病例和根治性手术后辅助外放疗的患者。以无复发生存(RFS)时间作为终点,采用多因素Cox回归分析法筛选放疗特异性候选基因。以基因回归系数和表达值乘积之和作为权重放疗敏感指数。应用时间依赖受试者工作特征(ROC)曲线分析该模型对RFS预测的准确性。应用Kaplan-Meier生存曲线联合Log-rank检验分析不同放疗敏感组间RFS的差异。结果数据库中子宫颈癌临床表型数据集中行根治性手术切除病例155例,其中完成辅助外放疗18例,转录组和微RNA(miRNA)数据库探针过滤后分别有17 156个基因和480个miRNA纳入分析。最终有23个基因符合统计学标准,经分层聚类分析这些基因为7个聚类,最终确立MAT2A、PIGY、SUSD1、ZNF341、TMEM85、HIF1AN和CLEC18A作为构建放疗敏感模型的基因,根据这7个基因构建放疗敏感指数,中位值为0.726(-5.501~9.046)。在单纯根治性手术组中,对脉管内癌栓校正后,7个基因敏感指数依然是RFS独立预后因素(HR=1.456,95%CI 1.192~1.780,P<0.01);在根治性手术后辅助放疗组中,对盆腔淋巴结阳性数校正后,7个基因敏感指数依然是RFS独立预后因素(HR=0.572,95%CI 0.328~0.998,P=0.049)。时间依赖性ROC曲线分析表明,7个基因放疗敏感指数均优于N分期及脉管内癌栓对3、5年RFS率预测的准确性。结论依据基因表达数据构建的子宫颈癌辅助放疗敏感模型具有判断术后放疗获益人群的潜在价值。

ObjectiveTo construct a radiosensitivity model based on gene expression in tumor tissues,and to explore the feasibility of using this model to select the patients benefiting from adjuvant radiotherapy in cervical cancer.MethodsPatients underwent radical hysterectomy alone or radical hysterectomy plus adjuvant radiotherapy were selected from The Cancer Genome Atlas(TCGA)cervical cancer dataset.Candidate radiosensitive genes were identified through multivariate Cox regression analysis with recurrence free survival(RFS)as end-point events.The sum of regression coefficients and products of expression levels of candidate genes was calculated as radiosensitivity index.Time-dependent receiver operating characteristic(ROC)curve analysis was used to analyze the predictive accuracy of this model for RFS.Kaplan-Meier plot coupled with Log-rank test was used to evaluate difference in RFS between subgroups with different radiosensitivity.ResultsA total of 155 patients underwent radical hysterectomy was retrieved from the phenotype dataset among which 18 patients was completely treated with planned adjuvant radiotherapy.Expression values of 17 156 genes and 480 microRNAs after probes filtration were included into analysis.Twenty-three genes were identified consistent with statistical criterion.Hierarchical cluster analysis revealed 7 subgroups existed across these 23 genes.The MAT2A,PIGY,SUSD1,ZNF341,TMEM85,HIF1AN and CLEC18A genes were finally selected to calculate radiosensitivity index.The median value and range of RI was 0.7258(-5.501-9.046).In subset consist of patients underwent radical hysterectomy alone,adjusted for intravascular tumor thrombus,radiosensitivity index was independent prognostic factor(HR=1.456,95%CI 1.192-1.780,P<0.01).Whereas,radiosensitivity index was remained significantly associated with RFS after adjustment for the number of positive nodes in subset comprised of patients underwent radical hysterectomy plus adjuvant radiotherapy(HR=0.572,95%CI 0.328-0.998,P=0.049).Moreover,radiosensitivity index exhibited more accuracy in prediction for 3 or 5 years RFS compared with clinical prognostic factors N stage and intravascular tumor thrombus.ConclusionThe radiosensitivity model based on gene expression data in tumor tissues has potential value to select the cervical cancer patients benefiting from adjuvant radiotherapy.