急性髓系白血病血清标志物初步筛选及临床意义

Preliminarily Screening of Serum Characteristic Markers in Acute Myeloid Leukemia and Clinical Significance

本研究初步分析初诊急性髓系白血病(acute myeloid leukemia,AML)患者与正常人群血清蛋白质质谱的差异,筛选AML血清特征性肿瘤标志物及探讨AML患者血清蛋白质组学特征及在白血病发病机制中的意义.选择AML患者14例和健康对照28例,应用弱阳离子交换纳米磁珠捕获血清蛋白质组分及Autoflex II基质辅助激光解吸电离飞行时间质谱仪检测各样品的蛋白质质谱,再利用CliprotoolsTM2.2软件进行数据分析,筛选差异蛋白质分子并建立AML诊断模型.选取验证组7例AML血清和14例健康对照血清对产生的模型用盲法进行验证.结果表明,在分子量700-10 000 Da范围内,可检测到69个左右的蛋白质谱峰.与健康组比较,AML患者组中共检测出有显著意义的蛋白峰差44个(p<0.0001),其中10个蛋白峰高表达,34个蛋白峰低表达.将区分疾病组与对照组能力最强的3个质谱峰建成一个基于QC(Quick Classifier Algorithm)算法的诊断模型,质荷比(mass charge ratio,m/z)分别为3216.57,4089.7,7762.87,在该训练集中其判断AML的预期敏感性为86.4%,预期特异性为82.8%.分类验证中该模型能正确识别7例AML病例中的6例和14例健康对照中的12例.交叉验证显示,该模型的敏感性和特异性均为85.7%.结论:初步应用Clinprot系统建立的急性髓系白血病QC模型由3个显著差异蛋白质峰构成,似能有效区分AML患者与正常对照者,其敏感性和特异性较高,有可能作为AML的血清标记物,其中m/z7762.87为血小板源性的趋化因子,其有可能为AML发病机制、分子分型、疗效及预后判断提供重要的实验依据.

This study was purposed to preliminarily screen characteristic tumor markers of acute myeloid leukemia（AML） and to investigate the serum proteomics characteristics of patients with AML and their significance in pathogenesis.14 patients with AML and 28 healthy controls were enrolled in this study.The serum protein components were captured by weak cation exchange nanometer magnetic beads,the protein mass-spectra of all samples were detected by Autoflex II matrix-assisted laser desorption/ionization time of flight mass spectrometer,and the detection data were analyzed by means of CliprotoolsTM2.2 software,then the differential protein molecules were screened and the diagnostic model was established.Sera of 7 AML patients and 14 healthy controls were selected to verify the established model by using blind test.The results indicated that about 69 protein peaks could be detected within the range of 0.7-10 kD in protein spectra of serum samples from AML patients and controls.Compared with healthy controls,there were 44 statistically differential expression peaks in AML group（p0.0001）.Among them,10 protein peaks were upregulated protein peaks and 34 protein peaks were downregulated.Diagnostic model was established on the basis of Quick Classifier Algorithm（QC）,and the three mass peaks had the strongest power for software to automatically distinguish AML group from control group.Mass charge ratios（m/z） were 3216.57,4089.7,and 7762.87 respectively.Sensitivity was expected as 86.4% while 82.8% in this established model group.Category validation showed that this diagnostic model correctly identified all 6 cases out of AML and 12 cases out of 14 healthy controls.In cross validation,the model sensitivity and specificity both were 85.7%.It is concluded that the AML QC model is composed of three protein peaks,which can effectively distinguish AML patients from healthy controls.Owing to higher sensitivity and specificity,they may act as serum tumor markers of AML.Among the three proteins,the one with m/z 7762.87 is the platelet-derived protein chemokine（PF4） protein.This finding will probably provide significant experimental evidence for understanding pathogenesis,molecular type,prognosis and treatment effect of AML.