利用蛋白质组学与人工神经网络构建大肠癌肝转移诊断模型

A Diagnostic Model for Colorectal Cancer Liver Metastasis by Proteomics and Artificial Neural Network

目的应用双向电泳联合质谱技术筛选大肠癌肝转移血清特异性标志物,并利用这些标志物以数据挖掘技术构建人工神经网络大肠癌肝转移诊断模型。方法收集大肠癌无肝转移、伴肝转移患者各12例血清样本,同组血清等量混合进行双向电泳,用ImageMaster V5.0软件分析两组蛋白质的差异,差异蛋白质进行MALDI-TOF-MS鉴定。ELISA法测定差异蛋白及CEA含量。用受试者工作曲线评价其对大肠癌肝转移的诊断价值,以人工神经网络方法,筛选出准确度最高者作为大肠癌肝转移诊断模型。结果双向电泳显示,大肠癌伴肝转移组与无肝转移组相比较有4个蛋白有统计学意义,其中2个上调的蛋白质分别是Transferrin和Complement component C9;2个下调的蛋白质分别是Haptoglobin和Isoform 1 of Serum albumin。受试者工作曲线分析与大肠癌肝转移相关性依次为Transferrin>Haptoglobin>CEA。人工神经网络的诊断方法以Transferrin和Haptoglobin这两个蛋白联合建立的模型预测准确度最高,为88.57%,其敏感度为63.64%,特异度为100%。结论利用蛋白质组学与人工神经网络构建了Transferrin与Haptoglobin联合的大肠癌肝转移诊断模型,横断面验证有较高的准确度,开辟了诊断大肠癌肝转移的新方法。

Objective To screen the specific proteins in serum between colorectal cancer pa- tients with liver metastases and without liver metastases by using two dimension electrophore- sis combined with the technology of mass spectrometry, and set up a diagnostic model for colorectal cancer liver metastases through the artificial neural network （ANN） method.Methods The serum was sampled from 12 colorectal cancer patients with liver metastases （CRCLM） and 12 colorectal cancer patients without liver metastases （CRC）. The concentrated and desal- ted samples were separated with two dimensional gel electrophoresis （2-DE） in triplicate ex- periments. The gels were scanned by Image Scanner after stained with Coomassie brilliant blue G-250. The biological variations of the protein expression level were analyzed with ImageMaster VS.0 software. The differentially expressed protein spots were identified by matrix- assisted laser desorption/ionization time of flight mass spectrometry （MALDI-TOF MS）. Transferrin and Haptoglobin were detected by ELISA and CEA was measured by automatic bio- chemistry analyzer in serum between the two group of patients. Receiver operating characteris- tic curve （ROC） was applied to evaluate these three proteins in classifying the two group of patients. The protein was gradually added to set up an ANN model to diagnose CRCLM based on the area under the curve（AUC）.The one with the most accuracy was selected as the diag- nostic model for colorectal cancer liver metastases. Results A total of 2 proteins identified as Transferrin and Complement component C9 were found to be significantly increased, and 2 i- dentified to be Haptoglobin and Isoform 1 of Serum albumin significantly decreased in the ser- um samples of colorectal cancer patients with liver metastases.The identified proteins included Haptoglobin, Complement component C9, Isoform 1 of Serum albumin and Transferrin. The or- der was Transferrin, Haptoglobin,CEA by ROC analysis. The ANN model consists of Transferrin and Haptoglobin kept the supreme accuracy to diagnose colorectal cancer liver metastases with 88.57% accuracy, 63.64% sensitivity and 100% specificity, and was selected to be the diagnosis model. Conclusion We can set up an artificial neural network model combining Transferrin and Haptoglobin with higher accuracy to diagnose colorectal cancer liver metastases.