Prediction Model for Diagnosis of Kawasaki Disease Using iTRAQ-Based Analysis

A quick prediction method may help confirm the diagnosis of Kawasaki disease (KD), and reduce the risk of coronary artery lesions. The purpose of this study was to evaluate potential candidate diagnostic serum proteins in KD using isobaric tagging for relative and absolute quantification (iTRAQ) gel-free proteomics. Ninety two subjects, including 68 KD patients (1.6 ± 1.2 years, M/F 36/32) and 24 fever controls with evident respiratory tract infection (2.1 ± 1.2 years, M/F 13/11) were enrolled. Medical records were reviewed for demographic and laboratory data. The iTRAQ gel-free proteomics was used to screen serum proteins completely and compare the difference between two groups followed by specific validation with ELISA. The candidate proteins and conventional laboratory items were selected for the prediction model of KD diagnosis by support vector machine. Five selected candidate proteins, including protein S100-A8, protein S100-A9, protein S100-A12, neutrophil defensin 1, and alpha-1-acid glycoprotein 1 were identified for developing the prediction model of KD diagnosis. They were used to develop an efficient KD prediction model with an area under receiver operating characteristic (auROC) value of 0.92 (95% confidence interval: 0.84, 0.98). These protein biomarkers were significantly correlated with the conventional laboratory items as follows: C-reactive protein, glutamic pyruvic transaminase, white blood count, platelet, segment and hemoglobin. These conventional laboratory items were used to develop a prediction model of KD diagnosis with an auROC value of 0.88 (95% confidence interval: 0.80, 0.96). Our result demonstrated that the prediction model with combined five selected candidate protein levels may be a good diagnostic tool of KD. Further prediction model with combined six conventional laboratory data is also an acceptable alternative method for KD diagnosis.