代谢组学分析技术筛选结核性胸膜炎患者胸腔积液生物标志物的研究

Discovery of the biomarkers from tuberculosis pleural effusion by metabolomic analytical techniques

目的采用超高效液相色谱．质谱联用分析技术（UPLC—MS）分析结核性胸膜炎患者胸腔积液，建立基于代谢物组的OPLS-DA模型，探讨潜在的代谢标志物。方法收集2012年11月至2013年9月在天津市海河医院住院患者的胸腔积液标本166份（结核性胸腔积液83份，细菌性胸膜炎31份，肺癌30份，心力衰竭22份）进行代谢组学定量分析研究。通过模式识别方法构建正交偏最小二乘判别分析（OPLS—DA）模型，代谢物根据变量投影重要性值（VIP）值、VIP置信区间进行初步筛选，然后利用SPSS17．0对所得变量进行非参数检验（Kruskal-WallisH检验），进而筛选出潜在代谢标志物。结果基于胸腔积液代谢物数据建立的OPLS．DA模型，验证模型的预测准确度达到100％（38／38），该模型能够很好地区分结核性胸膜炎组及对照组。基于代谢物筛查得到的46特征离子中有5个具有统计学差异，其中在结核性胸膜炎组17a，20a-二羟基胆甾醇（2589490．00），磷脂[20：4（8Z，11Z，14z，17z）]（1188670．00），生育三烯酚（1051760．00），磷脂（0—18：O）（434394．oo）与肺癌组（735615．00、336815．00、324563．00、193055．00）、细菌性胸膜炎组（1678805．00、598256．50、699384．00、343866．00）、心力衰竭组（535842．00、253503．00、234503．00、130185．00）比较显著增高（H分别为26．787、18．680、26，193、21．024，P〈0．01），在结核性胸膜炎组L-苯丙氨酸（245976．00）与肺癌组（753033．50）、细菌性胸膜炎组（357278．00）、心力衰竭组（586678．00）显著下降（H=13．635，P〈0．01）。结论基于UPLC-MS分析技术平台构建的OPLS-DA模型可区分结核性胸膜炎与对照组，为寻找结核性胸膜炎的特征标志物及早期诊断提供了新的方法。

Objective Pleura1 effusion of patients with tuberculous pleurisy was analyzed by ultra high performance liquid chromatography-mass spectrometry （UPLC-MS）. Orthogonal partial least squares discriminant analysis （OPLS-DA） model was established for searching and analyzing the potential metabolic biomarkers to provide new ideas for the early diagnosis of tuberculosis pleurisy, Methods Totally 166 cases of pleural samples were collected from November 2012 to September 2013 in Tianjin Haihe Hospital （tuberculosis pleurisy 83 cases, bacterial pleurisy 31 cases, lung cancer 30 cases and heart failure 22 cases） and metabonomics quantitative analysis was conducted. Quantitative analysis of metabolic methods was enrolled. Orthogonal partial least squares discriminant analysis （OPLS-DA） model was constructed by the pattern recognition method. Based on the OPLS-DA model, potential biomarkers was filtered preliminary by variable importance in the projection （VIP） and VIP confidence interval value. The specific metabolites were determined by applying non-parametric test（ Kruskal-Wallis H test）by using SPSS 17.0 , and potential metabolic biomarkers were screened. Results The prediction accuracy of OPLS-DA model was 100% （38/ 38）, which illustrated that the model could verify the tuberculous pleurisy group and the control group accurately. Based on the data of metabolites, 46 potential metabolites were finally screened and 5 metabolites were identified with statistically significant differences （ P 〈 0.05 ）. The data of tuberculosis pleurisy group showed a significant increase in 17a, 20a- Dihydroxy cholesteryl, phospholipid [20：4 （8Z, 11Z, 14z, 17Z） ] （ 1 188 670. 00）, tocotrienols （ 1 051 760. 00） and phospholipid（ O-18：0） （434 394.00） compared with the lung cancer group（735 615.00,336 815.00,324 563.00,193 055.00）, bacterial pleurisy group （1 678 805. 00,598 256. 50,699 384. 00,343 866. 00）, and heart failure group （535 842. 00,253 503.00, 234 503.00,130 185.00） （H = 26. 787,18. 680,26. 193,21. 024, P 〈 0. 01 ）, and a significant decrease in L- phenylalanine（245 976. 00）compared with the lung cancer group（753 033.50）, bacterial pleurisy group （357 278.00）, and heart failure group （586 678.00） （ H = 13. 635, P 〈 0.01 ）. Conclusions The OPLS- DA model constructed on the basic of UPLC-MS technology platform can verify the tuberculous pleurisy group and the control group accurately, and the study provides new ideas and methods for identifying features of tuberculous pleurisy markers and early diagnosis.