Background: Detection of hepatocellular carcinoma (HCC) in patients with chronic liver disease (CLD) is difficult. We investigated the use of comprehensive proteomic profiling of sera to differentiate HCC from CLD.
Methods: Proteomes in sera from 20 CLD patients with alpha-fetoprotein (AFP) <500 microg/L (control group) and 38 HCC patients (disease group) were profiled by anion-exchange fractionation (first dimension), two types (IMAC3 copper and WCX2) of ProteinChip Arrays (second dimension), and time-of-flight mass spectrometry (third dimension). Bioinformatic tests were used to identify tumor-specific proteomic features and to estimate the values of the tumor-specific proteomic features in the diagnosis of HCC. Cross-validation was performed, and we also validated the models with pooled sera from the control and disease groups, serum from a CLD patient with AFP >500 microg/L, and postoperative sera from two HCC patients.
Results: Among 2384 common serum proteomic features, 250 were significantly different between the HCC and CLD cases. Two-way hierarchical clustering differentiated HCC and CLD cases. Most HCC cases with advanced disease were clustered together and formed two subgroups that contained significantly more cases with lymph node invasion or distant metastasis. For differentiation of HCC and CLD by an artificial network (ANN), the area under the ROC curve was 0.91 (95% confidence interval, 0.82-1.01; P <0.0005) for all cases and 0.954 (95% confidence interval, 0.881-1.027; P <0.0005) for cases with nondiagnostic serum AFP (<500 microg/L). At a specificity of 90%, the sensitivity was 92%. Both cluster analysis and ANN correctly classified the pooled serum samples, the CLD serum sample with increased AFP, and the HCC patient in complete remission.
Conclusion: Tumor-specific proteomic signatures may be useful for detection and classification of hepatocellular cancers.