食管癌血清蛋白质谱揩纹图分析及人工神经网络诊断模型研究

Serum Protein Fingergrinting Spectral Analysis of Esophageal Cancer and Research for Artificial Neural Network Diagnosis Model

目的利用高通量和高灵敏的表面激光解析电离飞行时间质涪技术（SELDI—TOF—MS）寻找食管癌患者血清中微量的标志性差异蛋白质组，为探索食管癌发生与基因转录的蛋白调控机制，基因治疗以及早期诊断提供参考数据，方法用SELDI—TOF—MS检测食管癌及其相关人群的血清蛋白质谱指纹图，用Biomarkex Wizaid Software软件筛选出差异蛋白，选择具有标志性的差异蛋白质组建立人工神经网络诊断模型，使用SPSS分析其诊断效能。结果在食管癌患者血清中发现89个差异蛋白（P〈0．05）。其中有显著差异（P〈0.001）的蛋白质如下：在食管癌患者血清中表达增高的蛋白（5017．6Da，7458、5Da，7908，1Da，8111.9Da，8577，8Da）和表达降低的蛋白（4215.8Da，5890.9Da，7749.3Da）。利用筛选出8个有明显表达差并的蛋白质组建立食管癌人工神经网络筛查模型和诊断模型，其灵敏度达到93．2％和96.3％，特异度分别为95．6％和97．2％，经大样本盲法验证的灵敏度为75.4％和75.8％，特异度分别为84．8%和86．7％，经蛋白组数据库检索发现两种蛋白分别为血清淀粉样蛋白A和子宫球蛋白。结论血清蛋白质谱指纹图结合人工神经网络技术进行蛋白组学数据挖掘对食管癌的筛查和诊断以及探索其基因和蛋白调控机制具有重要的临床意义。

Objective The high-throughpat and highly sensitive surface enhanced laser desorptioa/ionization time of flight mass spectrometry （SELDI-TOF-MS） were used to look for a sign of trace differences in proteome of patients with esophageal cancer and explore the occurrence of esophageat cancer,gene transcription and protein regulatory mechanism. It could provide reference data for gene therapy and early diagnosis. Method：Using sermn protein fingsrprm. spectrum detesting by SELDI-TOF-MS for esophageal cancer and its associated groups in serum protein fingerpriut spectrum,and filtering out differences in protein group by Biomarker Wizaid Software which were identified by the correlate methods. Then selected symbolic differences in proteome to establish artificial neural network diagnosis rondel ,and finally analyse of its diagnostic performance by SPSS. Results Found 89 different proteins in esophagcal cancer patients （P〈0.05）. The particular significance of these proteins were as follow：some of them were very high intensity in the serum of patients with esophageal cancer such as peaks at 5017. 6,7458.5,7908. 1,8111.9 and 8577. 8Da. Instead,some were low inteqsity expression in peaks at 4215.8,5890.9 and 7749.3Da. Screcned out eight significant differences in protein xprestion in esophageal cancer group to establish the screening model and diagnosis model of artificial neural network, the sensitivity reached 93.2% and 96. 3% ,specificity was 95.6% and 97.2%. The two models were validated with a test set dauble blindly by the large sample validation. Their sensitivities were 75.4 % and 75.8% and specificities were 84.8 % and 86.7 %. Two of them were identified as serum Amyloid protein-A and Uteroglobin. Conclusion Serum protein fingerprint spectrum with artificial neural network technology to proteomics data mimng on the screening and diagnosis of esoohageal cancer,as well as explore the mechanism of gene and protein regulation have important clinical significance.