摘要目的 结合代谢组和计算生物学技术,探索食管癌代谢表型与病理分型的关系,发现食管癌患者代谢网络扰动机制,建立可用于术前快速对肿瘤进行精准分期的方法.方法 前瞻性队列研究设计,研究起止时间为2013年4月至2016年1月.纳入2013年5月至2014年4月在四川省人民医院确诊食管癌拟行手术患者,收集血清标本进行磁共振波谱代谢组学检测,绘制不同分期患者的血浆代谢指纹图谱.采用主成分分析、偏最小二乘法和支持向量机进行数据处理,并基于基因本体论(G0)方法建立代谢型-酶学网络-GO通路回溯分析技术探索调控食管癌代谢网络异常的酶-基因网络调节机制,最终建立食管癌分期定量预测模型.所有数据处理过程均在高性能计算平台上使用Matlab完成.计量资料的组间比较采用Wilcoxon秩和检验(非正态分布)或方差分析(正态分布).计数资料采用x2检验或Fisher确切概率法.结果 纳入不同TNM分期食管癌患者20例,不同分期患者组间年龄无差异(F=1.086,P＞0.05),体重指数无差异(F=1.035,P＞0.05),肿瘤距门齿距离无差异(F=1.078,P＞0.05),其血浆磁共振波谱指纹图谱可特征性区分不同肿瘤分期.在不同TNM分期的患者中,其血浆代谢池中存在明显的代谢物差异:与ⅡB、ⅢA、Ⅳ期患者相比,ⅠB、ⅡA期患者血清中丁酮、乙醇胺、同型半胱氨酸、羟基丙酸、雌三醇浓度升高,而糖蛋白、肌酸、胆碱、异丁酸、丙氨酸、亮氨酸、缬氨酸水平降低.经过计算,最终筛选出4个代谢标志物:乙醇胺、羟基丙酸、同型半胱氨酸、雌三醇.GO分析表明,调控这4个关键代谢标志物的是54对酶和基因.以此为基础建立了基于磁共振波谱的食管癌TNM分期定量预测模型,交叉验证结果表明,该预测模型的均方根误为5.3,R2 =0.47(P =0.036),预测效果良好.结论 基于代谢组学和酶-基因调节网络分析的系统生物医学方法能定量描绘中晚期食管癌患者代谢网络的扰动,且通过血浆磁共振波谱检验,能成功进行术前快速的食管癌患者临床分期.

abstractsObjectives By combining the metabolomics and computational biology,to explore the relationship between metabolic phenotype and pathological stage in esophageal cancer patients,to find the mechanism of metabolic network disturbance and develop a new method for fast preoperative clinical staging.Methods A prospective cohort study (from April 2013 to January 2016) was conducted.The preoperative patients from Sichuan Provincial People's Hospital,who were diagnosed with esophageal cancer from May 2013 to April 2014 were included,and their serum samples were collected to detect 1H-nuclear magnetic resonance (NMR) metabolomics for the purpose of drawing the metabolic fingerprinting in different stages of patients with esophageal cancer.The data were processed with these methods-principal components analysis:partial least squares regression and support vector machine,for the exploration of the enzyme-gene network regulatory mechanism in abnormal esophageal cancer metabolic network regulation and to build the quantitative prediction model of esophageal cancer staging in the end.All data were processed on high-performance computing platforms Matalab.The comparison of data had used Wilcoxon test,variance analysis,x2 test and Fisher exact test.Results Twenty patients with different stages of esophageal cancer were included; and their serum metabolic fingerprinting could differentiate different tumor stages.There were no difference among the five teams in the age (F =1.086,P ＞ 0.05),the body mass index (F =1.035,P ＞ 0.05),the distance from the incisors to tumor (F =1.078,P ＞ 0.05).Among the patients with different TNM stages,there was a significant difference in plasma metabolome.Compared to Ⅱ B,Ⅲ A,Ⅳstage patients,increased levels of butanone,ethanol amine,homocysteine,hydroxy acids and estriol,together with decreased levels of glycoprotein,creatine,choline,isobutyricacid,alanine,leucine,valine,were observed in Ⅰ B,Ⅱ A stage patients.Four metabolic markers (ethanol amine,hydroxy-propionic acid,homocysteine and estriol) were eventually selected.gene ontology analysis showed that 54 enzymes and genes regulated the 4 key metabolic markers.The quantitative prediction model of esophageal cancer staging based on esophageal cancer NMR spectrum were established.Cross-validation results showed that the predicted effect was good (root mean square error =5.3,R2 =0.47,P =0.036).Conclusions The systems biology approaches based on metabolomics and enzyme-gene regulatory network analysis can be used to quantify the metabolic network disturbance of patients with advanced esophageal cancer,and to predict preoperative clinical staging of esophageal cancer patients by plasma NMR metabolomics.