基于支持向量机的血清蛋白质指纹图谱模型在甲状腺癌诊断中的应用研究

目的 检测甲状腺癌患者血清蛋白质，筛选特异的蛋白质标记物，构建用于甲状腺癌早期诊断的血清蛋白质指纹图谱模型。方法 应用表面增强激光解吸电离主飞行时间质谱（SELDITOF-MS）技术测定81例血清标本（其中甲状腺癌40例，甲状腺腺瘤9例，健康人32例）的蛋白质质谱，用随机抽取的66例标本（甲状腺癌32例，甲状腺腺瘤9例，健康人25例）作为训练组，应用支持向量机进行训练和交叉验证，建立甲状腺癌诊断模型。结果区分甲状腺癌和正常人的诊断模型经留一法交叉检验该模型敏感性87.5％，特异性80％，用15例未知血清经盲法测试其敏感性为100％，特异性为86％；区分甲状腺癌和甲状腺腺瘤的诊断模型经留一法交叉检验敏感性为96．8％，特异性为89％。区分乳头状甲状腺癌和其他病理类型的甲状腺癌的诊断模型对乳头状甲状腺癌的判别率为97％，对其他病理类型的甲状腺癌的判别率为71％。结论表面增强激光解吸电离主飞行时间质谱技术结合支持向量机建立甲状腺癌血清蛋白质指纹图谱模型为早期筛查及诊断甲状腺癌提供了一种特异性强、敏感性高的新方法，值得进一步研究和应用。

Application of serum protein fingerprint model and support vector machine in diagnosis of thyroid cancer

Objective To detect new biomarkers and to establish a serum protein fingerprint model for early detection and diagnosis of thyroid cancer.Methods The serum samples of 40 thyroid cancer patients, 9 thyroid adenoma patient, and 32 healthy individuals were randomly divided into 2 sets: training set (n=66, including 32 thyroid cancer patients, 9 thyroid adenoma patients, and 25 healthy individuals) and test set(n=15). The serum protein were bound to WCX2 chip and tested by surface enhanced laser desorption/ionization time of flight-mass spectrometry (SELDI-TOF-MS). The data of spectra were analyzed by support vector machine (SVM) to establish a diagnostic model. Results The detective model combined with 3 biomarkers could differentiate the serum of thyroid cancer from that of healthy individual with a specificity of 86% and a sensitivity of 100%. The diagnostic model combined with 3 biomarkers could differentiate thyroid cancer from thyroid adenoma with a specificity of 88.9% and a sensitivity of 96.9%. The positive predictive value to differentiate papillary thyroid carcinoma from the thyroid cancer of other types was 97%, and the positive predictive value of thyroid carcinoma of other pathological types was 71%. Conclusion The combination of SELDI with bioinformatics tools is a novel, effective, and highly specific and sensitive method for thyroid cancer detection and diagnosis.