血清多肽谱预测乳腺癌新辅助化疗疗效的应用

Serum proteomic profiling predicts the susceptibility of breast cancer to neoadjuvant chemotherapy

目的 用纳米磁珠结合基质辅助激光解析离子化飞行时间质谱（MALDI-TOF-MS）技术检测乳腺癌新辅助化疗患者治疗前血清蛋白指纹图谱,筛选有疗效预测价值的相关蛋白,并建立疗效预测模型,探讨其在预测新辅助化疗疗效中的应用价值.方法 用MALDI-TOF-MS技术检测50例临床分期Ⅱ～Ⅲ期的浸润性导管癌患者治疗前血清标本,获得血清蛋白指纹图谱,患者行新辅助化疗2～4个周期后,根据RECIST标准评价疗效,分为新辅助化疗有效组（CR＋PR,31例）和无效组（SD＋PD,19例）.用Biomarker Wizard软件分析比较两组间的血清蛋白图谱,找出差异蛋白,分别采用k最近邻分类器（KNN）和支持向量机（SVM）两种分类器对筛选出来的蛋白位点进行分类.建立疗效预测模型,并进行盲法验证.结果 在相对分子质量1000～15000范围内,共检测到145个蛋白峰,化疗有效组与无效组比较,有3个蛋白质峰差异有统计学意义（P＜0.01）,m/z分别为2651、3452、2176.使用KNN分类器,由9个蛋白质峰（m/z为：2651、3452、2176、1585、1682、1908、10700、3014、8426）构建的预测模型,在预测疗效的准确率上达到84%,敏感性为100%,特异性为56%.结论 应用MALDI-TOF-MS技术可以筛选出乳腺癌化疗敏感相关的血清蛋白指纹图谱.

Objective To analyze the serum proteomic patterns in the breast cancer patients using matrix-assisted laser desorption/ionization time of flight mass spectrometry （MALDI-TOF-MS） before neoadjuvant chemotherapy, build predictive model and evaluate its clinic significance. Methods Fifty patients with clinical stage Ⅱ -Ⅲ of invasive ductal carcinoma were included in this study. Serum samples were prospectively collected before 2-4 cycles of neoadjuvant chemotherapy, and were analyzed using MAL-DI-TOF-MS. According to the response evaluation criteria in solid tumors （ RECIST）, patients were divided into 2 groups： drug susceptible group （31 cases, CR ＋ PR） and drug resistant group （ 19 cases, SD ＋PD）. Biomarker Wizard software was used to detect protein peaks significantly different between these two groups. The rule was built using two different supervised classification algorithms： K-Nearest Neighbor Clustering （KNN） and Support Vector Machines （SVM）. The method with the highest accuracy was selected as the optimal predicting algorithm. Results 145 major protein peaks were detected at the molecular range of 1000 to 15 000, and 3 major protein peaks were detected significantly different between drug susceptible group and drug resistant group （ P 〈 0. 01 ）, with Mass/Charge （m/z） values being 2651,3452, 2176 respectively. In the validation set, the supervised classification with the KNN model correctly classified most tumor responses with an accuracy rate of 84%, and sensitivity of 100%, specificity of 56%. The predictive model consisted of 9 protein peaks at Mass/Charge（m/z） 2651,3452, 2176, 1585,1682, 1908, 10 700, 3014, 8426 respectively. Conclusion MALDI-TOF-MS technique could screen related proteomic fingerprints in estimating the therapeutic effect of neoadjuvant chemotherapy.