血清蛋白质谱结合人工神经网络在宫颈癌诊断中的应用

目的:采用蛋白质芯片表面增强激光解吸电离飞行时间质谱(SELDI-TOF-MS)技术检测宫颈癌病人血清蛋白指纹图谱,通过差异蛋白组学筛选特有的蛋白标记物。方法:应用SELDI-TOF-MS技术和WCX2(弱阳离子)芯片采集58例宫颈癌患者和57例健康人血清蛋白质指纹图谱,采用Biomarker Wizard软件筛选差异蛋白质组。将115例血清随机分为两组:以训练组30例宫颈癌患者和30例健康人建立人工神经网络(ANN)模型,以验证组28例宫颈癌患者和27例健康人血清标本用于模型的双盲法验证。结果:宫颈癌患者与对照组血清蛋白质指纹图谱有145个差异表达的蛋白质峰(P0.05),筛选出质荷比(M/Z)分别为5912、5642、8702、4320、6432的标志蛋白(P10-6),建立人工神经网络模型,其对宫颈癌的诊断敏感性为92.86%,特异性为88.89%,阳性预测值为89.66%,阴性预测值为92.31%。结论:特征蛋白在宫颈癌患者较正常人血清明显的高表达或低表达,可能对宫颈癌的早期诊断和治疗后随访具有重要的指导意义。

Application of serum protein profile combined with artificial neural network in diagnosis of cervical cancer

Objective:To detect the serum protein fingerprint electropherogram of patients with cervical cancer by SELDI-TOF-MS,and screen specific protein biomarkers with different proteomics.Methods:The serum protein fingerprint electropherogram from 58 patients with cervical cancer and 57 healthy individuals were collected by SELDI-TOF-MS and WCX2 (weak cation),different proteomics was screened with Biomarker Wizard software.115 serum samples were divided into two groups randomly:30 patients with cervical cancer and 30 healthy individuals were selected to establish artificial neural network model,28 patients with cervical cancer and 27 healthy individuals were selected for double blind method.Results:145 differentially expression protein peaks were found in serum protein fingerprint electropherogram in patients with cervical cancer and control group (P<0.05),the M/Zs of proteins were 5 912,5 642,8 702,4 320 and 6 432 (P<10-6),the sensitivity,specificity,positive predictive value and negative predictive value of artificial neural network in diagnosis of cervical cancer were 92.86%,88.89%,89.66% and 92.31%.Conclusion:Compared to healthy individuals,the high or low expression of specific proteins may play important and directive roles in early diagnosis of cervical cancer and follow-up after treatment.