决策树联合生物标志在肺癌辅助诊断中应用  

Application of decision tree combined with biomarkers in auxiliary diagnosis of lung cancer

作　　者：魏小玲[1] 谭善娟[2] 何其栋[1] 王威[1] 吴拥军[1] 王静[3] 吴逸明[1] 

机构地区：[1]郑州大学公共卫生学卫生毒理学教研室,河南郑州450001 [2]青岛市市立医院医院感染管理科 [3]郑州大学第一附属医院呼吸内科

出　　处：《中国公共卫生》2013年第10期1479-1482,共4页Chinese Journal of Public Health

基　　金：国家自然科学基金(30972457;81001239);河南省重大科技攻关项目(112102310102);河南省医学科技攻关计划项目(2011020082)

摘　　要：目的 检测p16、RASSF1A和脆性组氨酸三联体基因（FHIT）甲基化水平及外周血DNA端粒长度,建立并探讨判别分析与决策树2种分类模型在肺癌辅助诊断中的意义。方法 采用甲基化特异性PCR、实时荧光定量PCR法测定200名正常对照、200例肺癌患者外周血p16、RASSF1A和FHIT基因甲基化水平和DNA端粒长度,建立决策树、判别分析2种肺癌判别诊断模型。结果 肺癌组和对照组中p16、RASSF1A和FHIT 基因启动子甲基化水平（%）分别为0.59（0.16~4.50）与0.36（0.06~4.00）（P=0.008）、27.62（9.09~52.86）与17.17（3.86~50.87）（P=0.038）、3.33（1.86~6.40）与2.85（1.39~5.44）（P=0.002）;端粒长度分别为（0.93＆plusmn;0.32）和（1.16＆plusmn;0.57）（P〈0.001）,4项生物标志在2组间差异均有统计学意义;判别分析、决策树对预测集的预测准确度分别为64%、83%;ROC曲线下面积分别为0.640、0.830,差异有统计学意义（P〈0.05）。结论 数据挖掘工具建立的决策树模型判别诊断肺癌的效果优于判别分析。Objective To diagnose lung cancer by detections of p16,RASSF1A,fragile histidine traid（FHIT）genes promoter methylation status,and relative telomere length in peripheral blood DNA and to identify the significances of discrimination analysis and decision tree for auxiliury diagnosis of lung cancer.Methods The levels of p16,RASSF1A,FHIT genes promoter methylation,and relative telomere length in peripheral blood DNA of 200 healthy individuals and 200 patients with lung cancer were measured by SYBR green-based quantitative methylation-specific PCR and quantitative PCR,respectively,and then discrimination analysis and decision tree models were developed.Results The levels（95% confidence interval）of p16,RASSF1A,and FHIT genes promoter methylation of the lung cancer patients and healthy individuals were 0.59（0.16-4.50）and 0.36（0.06-4.00）（P=0.008）,27.62（9.09-52.86）and 17.17（3.86-50.87）（P=0.038）,and 3.33（1.86-6.40）and 2.85（1.39-5.44）（P=0.002）,respectively,and the relative telomere lengths were 0.93＆plusmn;0.32 and 1.16＆plusmn;0.57（P〈0.001）.There were statistically significant differences in the four biomarkers between the two groups.The accuracies of discrimination analysis and decision tree models were 64% and 83%,respectively and the areas under receiver operating curve were 0.640 and 0.830,with statistically significant differences between the two groups（P〈0.05 for all）.Conclusion The efficacy of decision tree model established with data mining tool is better than that of discrimination analysis in auxiliary diagnose of lung cancer auxiliary.

关 键 词：肺癌 生物标志 决策树 辅助诊断 

分 类 号：R734.2[医药卫生—肿瘤]