Application of SELDI-TOF-MS Coupled With an Artificial Neural Network Model to the Diagnosis of Pancreatic Cancer

Background: There are no satisfactory biomarkers for screening pancreatic cancer. The surface-enhanced laser desorption/ionization time-of-flight mass spectrometry (SELDI-TOF-MS) technique has been used to identify biomarkers for different types of cancers. A new and reliable SELDI proteomic method for diagnosing pancreatic cancer is needed. Methods: Four hundred and fifty-five serum samples were tested by SELDI-TOF-MS matching on a gold chip. Samples were assigned to 1 of 2 subsets according to collection order, viz., a training set, and a testing set. The training set was used to identify statistically significant peaks and to develop the artificial neural network (ANN) model for diagnosing pancreatic cancer. The testing set was used in a blind test to validate the diagnostic efficiency of the ANN model. Results: A total of 62 proteins that differed between patients and controls were identified (P<0.05). Nine of these proteins (P<0.01; m/z at 4218 Da, 4238 Da, 4264 Da, 4480 Da, 5805 Da, 5928 Da, 9033 Da, 9258 Da, and 9299 Da) were chosen to develop the ANN. The model was subjected to a blind test using the testing set for diagnosis of pancreatic cancer. Sensitivity and specificity were 66.67% and 95.98%, respectively, and the accuracy was 93.47%. Conclusion: These preliminary results suggest that patients with pancreatic cancer may have serum proteins that differ from healthy controls. The ANN is a new method for diagnosing and identifying pancreatic cancer.