Serum Biomarkers Selection and Diagnostic Prediction of Early Silicosis Patients Using Bayesian Network and Neural Network

Sera of 79 workers exposed to silica and 25 healthy controls were determined by matrix-assisted laser desorption ionization mass spectrometry (MALDI-TOF MS). Based on the minimum error Bayes decision theory, serum biomarkers of early silicosis patients were selected by making use of the global optimal ability of the genetic algorithm. Mass spectrometric peaks of 22 proteins were selected and used by artificial neural network (ANN) to establish a diagnostic model. A blinded test shows the ratios of correctness, sensitivity and specificity are 96.15%, 96.25% and 96%， respectively. Search results of tandem mass spectra against a protein database show that the 1 777 u mass spectrometric peak is identified as C3f, which is a fragment of complement C3. The 1 777 u mass spectrometric peak is significantly decreased in silicosis patients. The results indicate that C3f may be the potential biomarkers for the diagnosis of early stage of silicosis.