基于高通量测序计算免疫细胞评分构建子宫内膜癌诊断与预后模型

Construction of endometrial cancer diagnosis and prognosis model based on immune cell score calculated by high throughput sequencing data

目的基于高通量测序数据计算子宫内膜癌组织中免疫细胞评分,探讨免疫细胞浸润与子宫内膜癌诊断、预后的关系。方法数据来源于癌症和肿瘤基因图谱(TCGA)数据库,采用caret包将数据按8∶2分为训练集和验证集,通过xCell对免疫细胞类型进行识别和丰度分析,Logistic回归分析对肿瘤组织进行判别、建模,Cox回归进行生存分析,曲线下面积(AUC)用于模型灵敏度和特异度分析,生存曲线采用Log-rank检验。结果共纳入544例子宫内膜癌样本和23例癌旁正常组织样本,病例组与对照组基线资料差异无统计学意义,均衡可比,P>0.05。不同免疫细胞在肿瘤组织与瘤旁正常组织的分布不同,对总样本、训练集及验证集有较好的区分(P<0.05),训练集AUC为0.93(0.89~0.97),验证集AUC为0.90(0.78~1.00)。样本总体、训练集及验证集2、3、5年生存概率AUC点估计值均≥55.7%,样本总体和训练集免疫评分低者生存概率高(P<0.05),验证集差异无统计学意义,P>0.05。结论子宫内膜癌变引发高强度免疫细胞浸润,依赖免疫细胞评分的统计学模型可用于子宫内膜癌诊断和预后评价。

Objective To calculate immune cell score by high-throughput sequencing data,explore the relationships between immune cell infiltration and the diagnosis or prognosis of endometrial cancer.Methods Data were obtained from the Cancer Genome Atlas(TCGA),then randomly separated into training and validation(8∶2)sets by Caret package.xCell was used to get information of immune cell subtypes and its abundance.Logistic regression analysis was used for discrimination and construction of diagnosis model,and Cox regression analysis was used for survival analysis.The sensitivity and specificity of models were quantified by area under the curve(AUC).Log-rank test was used for the comparison of survival curves.Results Baseline characteristics of data were comparable between endometrial cancer patients(n=544)and normal controls(n=23,normal tissues were obtained from corresponding endometrial cancer patients,P>0.05).The distributions of immune cell subtypes were differed between endometrial cancer tissues and controls.The diagnosis model could classify endometrial cancer tissues and normal tissues well in total sets,training sets and validation sets(P<0.05,training sets AUC:0.93(0.89-0.97),validation sets AUC:0.90(0.78-1.00)).The predictive abilities of models at 2,3,5 years measured by AUC values were all≥55.7%in total sets,training sets and validation sets.Compared with patients in higher prognostic immune risk score(pIRS)group,patients in lower pIRS group had higher survival probabilities in total sets and training sets(P<0.05).Conclusion Endometrial carcinogenesis induces abundant immune cell infiltration,the diagnosis and prognosis models based on immune cell score can be used in diagnosis and evaluation of prognosis among endometrial cancer patients.