Zhao Haiyue, Sun Qian, Li Lisong, Zhou Jinhua, Zhang Cong, Hu Ting, Zhou Xuemei, Zhang Long, Wang Baiyu, Li Bo, Zhu Tao, Li Hong
Center of Reproduction and Genetics, The Affiliated Suzhou Hospital of Nanjing Medical University, Suzhou Municipal Hospital, Suzhou, 215002, China.
Cancer Biology Research Center (Key laboratory of the ministry of education), Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan 430030, China.
J Cancer. 2019 Jan 1;10(2):397-407. doi: 10.7150/jca.28127. eCollection 2019.
Primary platinum-based chemoresistance occurs in approximately one-third of patients with serous ovarian cancer (SOC); however, traditional clinical indicators are poor predictors of chemoresistance. So we aimed to identify novel genes as predictors of primary platinum-based chemoresistance. Gene expression microarray analyses were performed to identify the genes related to primary platinum resistance in SOC on two discovery datasets (GSE51373, GSE63885) and one validation dataset (TCGA). Univariate and multivariate analyses with logistic regression were performed to evaluate the predictive values of the genes for platinum resistance. Machine learning algorithms (linear kernel support vector machine and artificial neural network) were applied to build prediction models. Univariate and multivariate analyses with Cox proportional hazards regression and log-rank tests were used to assess the effects of these gene signatures for platinum resistance on prognosis in two independent datasets (GSE9891, GSE32062). AGGF1 and MFAP4 were found highly expressed in patients with platinum-resistant SOC and independently predicted platinum resistance. Platinum resistance prediction models based on these targets had robust predictive power (highest AUC: 0.8056, 95% CI: 0.6338-0.9773; lowest AUC: 0.7245, 95% CI: 0.6052-0.8438). An AGGF1- and MFAP4-centered protein interaction network was built, and hypothetical regulatory pathways were identified. Enrichment analysis indicated that aberrations of extracellular matrix may play important roles in platinum resistance in SOC. High AGGF1 and MFAP4 expression levels were also related to shorter recurrence-free and overall survival in patients with SOC after adjustment for other clinical variables. Therefore, AGGF1 and MFAP4 are potential predictive biomarkers for response to platinum-based chemotherapy and survival outcomes in SOC.
原发性铂类化疗耐药发生在大约三分之一的浆液性卵巢癌(SOC)患者中;然而,传统临床指标对化疗耐药的预测能力较差。因此,我们旨在鉴定新的基因作为原发性铂类化疗耐药的预测指标。进行基因表达微阵列分析,以在两个发现数据集(GSE51373、GSE63885)和一个验证数据集(TCGA)上鉴定与SOC中原发性铂耐药相关的基因。采用逻辑回归进行单因素和多因素分析,以评估这些基因对铂耐药的预测价值。应用机器学习算法(线性核支持向量机和人工神经网络)构建预测模型。采用Cox比例风险回归和对数秩检验进行单因素和多因素分析,以评估这些基因特征对铂耐药的影响在两个独立数据集(GSE9891、GSE32062)中的预后情况。发现AGGF1和MFAP4在铂耐药的SOC患者中高表达,并独立预测铂耐药。基于这些靶点的铂耐药预测模型具有强大的预测能力(最高AUC:0.8056,95%CI:0.6338 - 0.9773;最低AUC:0.7245,95%CI:0.6052 - 0.8438)。构建了以AGGF1和MFAP4为中心的蛋白质相互作用网络,并鉴定了假设的调控途径。富集分析表明,细胞外基质异常可能在SOC铂耐药中起重要作用。在调整其他临床变量后,高AGGF1和MFAP4表达水平也与SOC患者较短的无复发生存期和总生存期相关。因此,AGGF1和MFAP4是SOC中铂类化疗反应和生存结果的潜在预测生物标志物。