| Objective: To reseach the changes of serum protein profiling in eso-phageal cancer for screening specific protein markers and creating diag-nostic models.To provide new experimental evidence for personalized medicine such as early esophageal cancer screening, clinical diagnosis, staging and classification, recurrence or metastasis, prognosis estimation and so on .Methods: The serum samples which were composed of vari-ous clinical classifications and stages of cancer (such as esophageal squamous cell carcinoma and other tumors) identified by the correspond-ing pathological diagnostic criteria, other benign diseases and healthy controls were collected. Corresponding samples of serum protein finger-prints were detected by surface-enhanced laser desorption ionization time of flight mass spectrometry (SELDI-TOF-MS) and supporting the gold chip.Screening differentially expressed proteins by Biomaker Wizard 3.1 software from protein fingerprints. Model of biomarkers was constructed and evaluated by the blind test using artificial neural network (ANN) software.Then Performance analysis of models were evaluated by SPSS11.5. At last, identified though the protein databases and analyzed the differences in protein. Results: A total of 89 discriminating m/z peaks were identified that were related to esophageal cancer and its related peo-ple (P < 0.05). Eight proeins that the mass charge ratios were 4215.8,5017.6,5890.9,7458.5,7749.3,7908.1,8111.9,8577.8 had significant differences (P value <0.001)and discriminating value. The increased ex-pression in esophageal cancer were 5017.6,7458.5,7908.1,8111.9,8577.8. the rest (4215.8,5890.9,7749.3) were low expression. The Screening and diagnostic models of biomarkers constructed by ANN software based on the eight biomarkers generated excellent separation between the esophageal cancer and control groups. The sensitivity of models were respectively 93.2% and 96.3% and the specificity were 95.6% and 97.2%.After large sample of blind test, the sensitivity were 75.4% and 75.8% and the specificity were 84.8% and 86.7%. Searching the related databases, two of them were the Amyloid protein A and Ute-roglobin. Compared early esophageal spectra with advanced esophageal cancer and established the staging model by different proteins in different stages. Between esophageal cancer and controls, there were five differ-ences in proteins. Three proteins in early and advanced esophageal cancer were found. The five proteins were selected to build the model of early diagnosis with a sensitivity of 87.88%, a specificity of 91.43%, and an accuracy of 89.71%.The blind test generated a sensitivity of 95.83%, a specificity of 89.13% and an accuracy of 91.43%. Three different proteins were used to establish the model of staging diagnosis. The sensitivity was 75.76%, the specificity was 79.17%, and the accuracy was 77.19%. After analyse the Protein fingerprinting binding postoperative pathologic dif-ferentiation, five differential proteins were found in high, medium , low differentiation group of esophageal cancer and the healthy control group.The mass charge ratios (M/Z) of them were 4654.7,5948.2,6645.7,7734.2 and 9301.1 respectively. Using the proteins selected by artificial neural network software to creat predictive diagnos-tic model of esophageal cancer. The sensitivity and specificity of diag-nostic model were 81.25% and 85.38% and the accuracy was 83.47%. Conclusion: Combined fingerprint serum protein profiling with artificial neural network techniques in proteomics data mining to have important clinical significance .Because it was vital in screening and diagnosis of esophageal cancer, researching the pathogenesis of esophageal cancer, gene and protein regulation mechanisms.The diagnostic model of staging and classification opened up the the way that identificated the degree of differentiation of esophageal cancer cell in molecular level and perhaps could provide references for the molecular pathology of cancer cells...
|
PDF Full Text Request