蛋白指纹图谱技术在骨肉瘤临床诊断中的初步应用

Use of SELDI-TOF-MS method to identify patients with osteosarcoma

目的比较骨肉瘤患者和正常对照者血清蛋白表达谱的差异，筛选骨肉瘤相关血清蛋白标志物，并建立基于决策树的预测模型，为筛选和建立骨肉瘤临床诊断的血清学指标提供依据。方法27例骨肉瘤患者血清（男17例，女10例）及47名相匹配者正常对照血清标本随机分为两组：60份（23例骨肉瘤，37名正常对照）为建模组，14份（4例骨肉瘤，10名正常对照）为盲法筛选组。利用表面增强激光解吸离子化-飞行时间-质谱（surface enhanced laser desorption／ionization time of fight mass spectrometry，SELDI—TOF—MS）技术进行蛋白质谱分析。采用蛋白质飞行质谱仪对结合在CM10芯片上的血清蛋白进行读取分析。通过BiomarkerWizard软件比较两组人群血清蛋白质谱的差异，经生物信息学分析得到决策树模型并进行盲法验证。结果在质荷比（M／Z）1488．15-19842u范围内，共检测到96个有效蛋白峰，其中9个峰差异有统计学意义。利用三倍交叉证实方法对建模组的蛋白质谱数据进行1000次随机抽样，得到1000个决策树。根据交叉证实的正确率选出最佳的20个决策树模型作为最终预测模型。用其对14个盲法筛选样本进行归类预测的正确率为85．71％。结论应用SELDI—TOF—MS技术可筛选出骨肉瘤相关血清蛋白标志；建立的决策树模型可以对骨肉瘤作出较为准确的预测判断。

Objective To provide some theoretic evidence for screening and establishing serum indicators of early diagnosis of osteosarcoma（OS）, the serum proteomics profiling difference of subjects with OS and age-matched healthy controls were analyzed to screen serum proteomic biomarker related to osteosarco- ma. Methods Serum samples were collected from 27 patients of OS （17 males, 10 females） and 47 age- and sex-matched healthy controls. The samples were divided into 2 sets randomly： training set （23 OS pa- tients, 37 healthy controls） and blind testing set（40S patients, 10 healthy controls）. Special serum protein or peptide pattern was determined by SELDI-TOF-MS measurement after treating the sample onto CM10 pro- tein chip. The obtained data were analyzed by Biomarker Wizard software to screen serum proteome biomarkers with relation to OS, while decision tree for diagnosis of OS and blind validation were determined by bioinformatics analysis. Results 96 effective protein peaks were detected at the molecular range of 1488.15-19842 u, among which 9 were significantly different between OS and controls. All the peptide pat- tern data were sampled randomly 1000 time, and 1000 decision tree model were obtained. Decision tree and 3-cross validation approach were used combine, 20 decision tree which can differentiate effectively OS pa- tients from controls were constructed. With these classification tree, 12 samples were correctly forecasted in 14 blind testing samples, the corresponding correct rate was 85.71%. Conclusion SELDI-TOF-MS protein chip combined with artificial intelligence classification algorithm helps find serum proteome biomarkers relat- ed to OS and the predictive models can discriminate OS from healthy controls effectively, which may have some potential value for early diagnosis of OS.