| Aims:To investigate the difference of serological proteomics in hepatocellular carcinoma (HCC) compared to other liver diseases, and explore the high sensitive and specific diagnostic model and new biomarker for HCC. Methods:The serum and clinical characters of HCC (n=153), liver cirrhosis (LC) (n=95), hepatitis (n=115), intra-hepatic cholangio-carcinoma (ICC) (n=15), metastatic cancer from adenocarcinoma (MC) (n=30), focal nodular hyperplasia (FNH) (n=6), angioma (n=12) and healthy controls (n=109) were collected for clinical data set. The epidemiology and statistics analyzed the difference of laboratory parameters between HCC and other benign/malignant tumors; ELISA method tested candidate biomarker GP73 (Golgi protein73) in liver diseases; ten fold cross validation and linear SVM analyzed the model of ALT, AST, AFP and GP73; the serum was treated by MB-WCX beads and the proteomic profiling was attained by Clinprot/Matrix-Assisted Laser Desorption/Ionization Time of Flight Mass Spectrometry (MALDI-TOF-MS); the purified proteins was identified by FT-ICR-MS. Results:i After establishing the clinical data base,19 laboratory parameters in HCC groups were distinct with other groups; the model differentiating HCC from LC was set up with the sensitivity of 85.2% and specificity of 84.2%. ii Median sGP73 in LC was higher than in HCC and hepatitis (P=0.001), and sGP73 in all three groups were higher than those in healthy individuals (P<0.001); AFP/GP73 had a sensitivity of 75.8% and specificity of 79.7% with an area under the receiver operating curve (AUROC) of 0.84. Te positive rate of sGP73 in angioma, FNH, ICC and MC was 0,50,63.3,53.3%, respectively. iii Ten fold cross validation and linear SVM could combine the clinical parameters and raise the diagnostic ability. When differentiating disease and normal group, the average accuracy of ALT+AFP+GP73 was 85.35%; when differentiating hepatitis and LC/HCC, the average accuracy of ALT+AST+AFP+GP73 was 79.77%; and the accuracy of AST+AFP+GP73 was 88.26% when differentiating LC and HCC.â…³The specific model comprised of two peptides 2882.89 Da and 4476.12 Da could distinguish HBV infected (chronic hepatitis and asymptomatic carriers (AsC)) from healthy (HBV-immunized and normal) group showed 95.5% of sensitivity and 95.4% of specificity by cross-validation analysis.40/56 HBV infected and 43/50 healthy subjects could be correctly classified by the model. The peaks,2882.89 Da and 4476.12 Da, identified as fibrinogen beta chain (FBG) and nucleophosmin (NPM) respectively were both up to 0.88 when discriminating AsC from healthy group,â…´Little disparity of peptidomic profiling was found between AFP (+) and (-) HCC groups. m/z 3508 Da and 2954 Da, showed significant difference between AFP (-) HCC and LC groups (P<0.05); the area under the receiving operating curve of 3508 Da and 2954 Da was 0.87 and 0.82 respectively. SNN algorithm established the diagnostic model that was able to differentiate AFP negative HCC from LC group with recognitive rate at 86.37%, predictive rate at 70.29%.â…µThe total recognition rate of the diagnostic model differentiating HCC from ICC was up to 97.02%, the predictive rate was just 67.51%; the accuracy of the diagnostic mode between HCC and benign tumors, recognizing HCC was 98.59%, and benign tumors 64.29%; The total accuracy of the model between HCC and recurrence/metastasis was 92.59%, the predictive rate was 62.02%. Conclusions:Based on the clinical data base of liver diseases, the combination of bioinformatics and clinical data could set up simpler clinical diagnostic model; meanwhile, hepatic disease related biomarkers, such as GP73, FBG and NPM, were screened and will be further verified.
|
PDF Full Text Request