基于TCGA数据库建立的八基因预后模型在乳腺癌中的应用

目的 利用TCGA数据库建立预测乳腺癌预后的多基因预后模型， 分析多基因预后模型与乳腺癌各临床病理特征之间的关系。方法 对TCGA数据库中乳腺癌患者的mRNA数据进行整理， 通过R语言软件筛选出在乳腺癌样本及正常样本中差异表达的基因， 采用Cox比例风险回归模型从中筛选和建立多基因预后模型， 计算预后评分。根据预后评分的中位数将患者分为高风险组和低风险组。将临床病理因素和预后评分因素纳入Cox回归模型分析乳腺癌患者的生存影响因素。根据年龄、 ER受体状态、 HER-2表达情况、 淋巴结转移状态及病理分期进行分组， 采用Kaplan-Meier （K-M） 方法以多基因预后模型作为影响因素进行生存分析， 验证多基因模型对总体及各亚组乳腺癌患者中的预后价值。结果 将分析得到的2 142个差异基因纳入Cox回归分析， 共筛选出8个差异基因， 包括羧基酯脂肪酶 （CEL）、 POU区域转录因子 （POU3F2）、 维生素D-24羟化酶 （CYP24A1）、 脂肪酸结合蛋白7 （FABP7）、 MURC、 G 蛋白偶联受体 （GCCR）、 低密度脂蛋白受体相关蛋白-1B （LRP1B） 及丝氨酸蛋白酶2 （PRSS2）， 并建立八基因预后模型。预后评分 （PI） 公式为： PI=0.156×CEL的表达量+0.112×POU3F2的表达量-0.071×CYP24A1的表达量-0.065× FABP7的表达量+0.135×MURC的表达量-0.201×GCGR的表达量-0.063×LRP1B的表达量-0.090×PRSS2的表达量。计算709例患者预后评分后， 中位值为0.98， 共有355例患者纳入高风险组， 354例患者纳入低风险组。Cox回归分析显示， 年龄、 病理分期和八基因预后模型均是乳腺癌患者预后的独立影响因素 （P＜0.05）。生存分析证实， 在总体乳腺癌患者及各亚组 （除Ⅳ期外） 乳腺癌患者中， 预后评分低风险的患者总体生存率明显升高， 差异有统计学意义 （P＜ 0.01）。结论 八基因预后模型可用于预测乳腺癌患者的预后， 在根据临床病理特征分组的乳腺癌亚组中得到了验证， 有利于进一步指导临床治疗。

Identification of prognostic eight-gene signature model in breast cancer using integrated TCGA database

Objective To establish a multi-gene prognostic model for predicting the prognosis of breast cancer using Cancer Genome Atlas (TCGA) database, and to analyze the relationship between the multi-gene prognostic model and clinical and pathological features of breast cancer. Methods The mRNA data and clinical information of breast cancer cohort were downloaded from TCGA database. Differentially expressed genes (DEGs) were identified by R language software in breast cancer tissues and normal tissues. DEGs related to overall survival of patients were selected by univariate Cox regression model, and a multi-gene signature model was identified by multivariate Cox regression model. Patients were divided into high risk cohort and low risk cohort according to prognostic index calculated by prognostic index formula based on the result of multivariate Cox regression model. Factors were analyzed by univariate and multivariate Cox regression models according to clinicopathological characteristics and prognostic index related with survival of patients with breast cancer. Survival analysis of subgroups was conducted according to age, estrogen receptor status, Her-2 receptor status, lymph node status and pathological stage. Kaplan-Meier（K-M）survival analysis was used to evaluate the prognostic prediction of the multi-gene signature in overall patients and subgroups. Results Eight DEGs were selected to conduct a survival related multi-gene signature from total of 2 142 DEGs in univariate and multivariate Cox regression model analysis including CEL， POU3F2， CYP24A1， FABP7， MURC， GCCR， LRP1B and PRSS2. Prognostic index formula was as follows： PI=0.156 × the expression of CEL + 0.112 × the expression of POU3F2-0.071 × the expression of CYP24A1-0.065 × the expression of FABP7+0.135×the expression of MURC-0.201×the expression of GCGR-0.063×the expression of LRP1B- 0.090×the expression of PRSS2. Cox regression model analysis demonstrated that age, pathological stage and eight-gene signature were validated as the novel and independent prognostic factors (P＜0.05). According to survival analysis (K-M plot), the accurate prognostic performance of eight-gene signature was confirmed in both overall patients and subgroups (except Ⅳ stage). Patients with low risk of prognostic score showed significantly longer OS compared with patients of high risk of prognostic score (P＜0.01). Conclusion The eight-gene prognostic signature can be used to predict the prognosis of breast cancer patients. It is verified in the subgroup of breast cancer according to the clinicopathological features, which is helpful to further guide the clinical treatment.