肾癌血清蛋白质指纹图谱诊断模型的建立

Establishment of diagnostic model for renal cancer by serum protein fingerprinting

目的:筛选肾癌特异相关蛋白质,建立用于肾癌诊断的分类树模型,为进一步临床应用奠定基础。方法:采集肾癌患者48例与正常人83例血清,用表面增强激光解吸/电离飞行时间质谱分别检测其蛋白表达谱,用BioMarker Wizard软件筛选出差异蛋白,再用BioMarker Patterns软件建立肾癌诊断最优分类树模型。结果:在48例肾癌患者、83例正常人血清中共检测出44个差异蛋白质峰,在质荷比从2 800到16 100的差异蛋白中有21个蛋白质相对含量差异有统计学意义(P<0.05),从中选出12个差异蛋白建立分类树模型,用于鉴别肾癌患者与正常人,该模型在学习模式下的诊断准确率、灵敏度和特异性分别为98.47%(129/131),97.91%(47/48),98.79%(82/83),在测试模式下这3项指标分别为87.02%(114/131),85.42%(41/48),87.95%(73/83)。结论:肾癌血清蛋白质指纹图谱诊断模型具有一定优越性,为肾癌早期诊断提供了新途径。

Objective To explore the distinct expression of serum proteins of renal cancer and create a decision classification tree model for diagnozing renal cancer. Methods Serum samples from renal cancer patients and healthy people were collected and their serum protein fingerprinting were read by surface enhanced laser desorption and ionization time-of-flight mass spectrometry. The BioMarker Wizard software and BioMarker Pattern software were used to screen and analyze the distinct proteins between the two groups. The best classification tree model was created for diagnosing renal cancer. Results Forty-four protein features were stably detected in the serum protein pattern in 48 renal cancer patients and 83 healthy people, in which 21 distinct proteins of statistical significance were found at the mass-to-charge ratios range from 2 800 to 16 100 （P〈0. 05）, from which 12 distinct proteins were selected to create a decision classification tree model for distinguish renal cancer patients and healthy people. In the learning mode its accuracy of diagnostic was 98.47 % （129/131）, the sensitivity and specialty were 97.91% （47/48） and 98.79% （82/83）, respectively. In the testing mode the accuracy of diagnostic was 87.02% （114/131）, the sensitivity and specialty were 85.42% （41/48） and 87.95% （73/83）, respectively. Conclusion The diagnostic model of serum protein fingerprinting is useful and offers a new approach for the early diagnosis of renal cancer.