【作者】 黄海燕；

【导师】 敖露；

【作者基本信息】 福建医科大学 ， 生物信息学， 2020， 硕士

【摘要】 胰腺癌是一种预后极差的侵袭性疾病,其致死率位居恶性肿瘤第3位。胰腺导管腺癌(pancreatic ductal adenocarcinoma,PDAC)约占胰腺癌总数的90%。由于早期PDAC患者缺乏特异性症状,故仅有10%-15%的患者在初次临床就诊时可采取根治性手术切除伴随辅助治疗,但大部分患者术后肿瘤复发。目前常用的临床病理指标,如:TNM分期、糖类抗原19-9等均无法准确评估PDAC患者的预后。有研究者根据多个基因的表达值构建打分阈值对患者进行预后评估。然而基因表达的检测值极易受到实验批次效应及临床样本质量的影响。此外,这类标志应用时需预先收集样本进行数据标准化,限制其在临床方面的实际应用。样本内基因表达值的相对大小关系(relative expression orderings,REOs)对实验批次效应及临床样本质量的不确定性具有鲁棒性。本研究基于这一定性特征构建出能准确预测I和II期PDAC患者术后预后的基因对标志。首先,利用单因素COX风险比例模型,从ICGC数据库中161个I和II期PDAC患者的基因表达谱数据中筛选与患者整体生存(overall survival,OS)时间显著相关的基因,两两配成基因对后进一步筛选REOs模式与患者OS时间显著相关的基因对。根据前向序列搜索算法筛选出最优的基因对组合作为最终的预后基因对标志。将该标志应用于不同平台检测的独立验证数据集中,评估标志区分出的高、低风险组患者OS时间的差别。接着,在不同数据集的高、低风险组中筛选差异表达基因并进行功能富集分析。最后,比较不同预后组的患者在基因组和表观遗传组的特点。本研究开发出由64个基因对(64-GPS)组成的标志来评估I和II期PDAC患者的预后。至少有25个基因对提示患者处于高风险时,患者被归为高风险组。在来自不同实验室的六组独立数据集共324个患者中,高、低风险组患者的OS时间存在显著差异。低风险组中的患者的OS时间显著高于高风险组的患者。多因素Cox比例风险回归分析显示,在校正与患者OS时间相关的临床病理指标如淋巴结、肿瘤大小、分级、年龄和性别后,64-GPS仍与患者的OS时间显著相关。转录组分析表明,在所有数据集中,由TNM分期系统定义的转移和非转移组患者间的差异表达信号较弱且不可重复,而由64-GPS重新分类的高风险和低风险组患者间的差异表达信号显著一致且高度可重复。表观组学分析表明,基因启动子区域的高甲基化模式可能会导致高、低风险组患者间转录差异。基因组学分析表明,LRMDA,MAPK10和CREBBP等基因突变可能与PDAC患者的不良预后密切相关。综上所述,64-GPS可准确地评估I和II期PDAC患者的预后,为临床个体化治疗提供了理论依据,对患者治疗方案的制定具有指导意义。更多还原

【Abstract】 Pancreatic cancer is an aggressive disease with a very poor prognosis and is the third leading cause of cancer death.Approximately 90% of pancreatic cancer is pancreatic ductal adenocarcinoma(PDAC).Only 10-15% of PDAC patients can be treated with radical surgical resection and adjuvant therapy at the initial diagnosis because of atypical and unspecific symptoms in the early stage of tumor,but most patients have tumor recurrence after surgery.The currently used clinicopathological factors,such as TNM staging system and carbohydrate antigen 19-9,cannot accurately evaluate the prognosis of PDAC patients.Some researchers have constructed score thresholds based on the expression measurements of multiple genes to evaluate the prognosis of patients.However,the gene expression measurements are extremely susceptible to the experimental batch effects and the quality of clinical samples.In addition,the application of such signature requires pre-collection of samples for data normalization,which is impractical in clinical.The within-sample relative expression orderings(REOs)of gene pairs is robust to experimental batch effects and quality uncertainties of clinical samples.Based on this qualitative transcriptional feature,we can develop a gene pair signature that can accurately predict the prognosis of stage I and II PDAC patients after surgery.First,according to the univariate Cox proportional hazards regression model,gene expression profiles of 161 stage I-II PDAC patients from the International Cancer Genome Consortium database are used to identify candidate genes that are significantly associated with the overall survival(OS)time of patient.Then,for every two of the candidate genes,the gene pair whose REOs pattern significantly associated with the patient’s OS time are selected.A forward selection procedure is used to search for the optimal subset with the highest C-index value,which is used as the final prognostic gene pairs signature.The signature is applied in other independent validation datasets produced by different laboratories to assess the difference in patient’s OS time between the high-and low-risk groups.Differentially expressed genes between high-and low-risk groups in different datasets are identified and functional enrichment analysis is performed.Finally,the characteristics in genomic and epigenetic between the two prognostic subgroups are compared.In this study,a signature consisting of 64 gene pairs(64-GPS)was developed to predict the prognosis of stage I and II PDAC patients.Patients were divided into the high-risk group when at least 25 of 64 gene pairs suggested this patient was at high risk.In 324 samples from six independent validation datasets produced by different laboratories,the patient’s OS time was significantly different between the high-and low-risk groups.All patients in the low-risk group had significantly better OS time than patients in the high-risk group.The multivariate Cox regression analysis showed that the 64-GPS remained significantly associated with the patient’s OS time after adjusting available clinicopathologic factors,such as node,tumor size,grade,age and gender.Transcriptomic analysis showed that the differential expression signals between the metastatic and non-metastatic groups defined by the TNM staging system were weak and irreproducible in all datasets,whereas the differential expression signals between the high-and low-risk groups reclassified by 64-GPS were significantly consistent and highly reproducible.The epigenomic analysis showed that the hypermethylation pattern in the promoter region may lead to transcriptional change between the high-and low-risk groups.The genomic analysis revealed that mutation of genes such as LRMDA,MAPK10 and CREBBP may be related to the poor prognosis of PDAC patients.Conclusively,the 64-GPS can robustly predict the prognosis of patients with stage I-II PDAC,which provides theoretical basis for clinical individualized treatment and has guiding significance for the formulation of treatment plan of patients.更多还原