| Hepatocellular carcinoma (HCC) is among the most prevalent malignant tumor types, and bothmorbidity and mortality of HCC are in the forefront of the malignant tumors. The currentlyradical treatment is surgical resection, while the high rate of postsurgical recurrence remains aserious threat to patients’ survival. To data, there is no effective molecular model to predict theprognosis of HCC patients yet, which makes it difficult to choose proper adjuvant therapies. Thisthesis aims to build a discriminant analysis model that can accurately distinguish patients withdifferent prognosis and guide the clinical treatment, so that the patients at high risk of recurrencecan receive more effective adjuvant therapies while the low-risk patients will not be over-treated.Based on the expression pattern of49molecular markers in paired HCC and adjacentnon-tumor tissues from336patients who received curative resection, we found that multiplemolecular markers reflecting the immune microenvironment and biological characteristics oftumors were significantly associated with HCC recurrence. Then we made a furthercomprehensive comparison of several commonly used statistical classification model, improvedcommonly used feature selection methods, and finally constructed an SVM classification modelbased on a nonlinear kernal which consists of eight molecular markers, including B7-DCTã€Bcl-2_Tã€CD80_Tã€CK19_Tã€Cyclin D1_Tã€FasL_Pã€HLA-DR_Pand Ki-67_T. This model could achievea high accuracy in both cross-validation and independent validation group. Furthermore, thisclassifier could be used as an independent predictor of HCC recurrence in the multivariateanalysis considering commonly used indicators for discrimination(P <0.0001).In summary, the in situ molecular classifier could efficiently identify patients at higherrisk for postsurgical recurrence of HCC, and could provide a new direction for systematic and improved clinical follow-up or chemopreventive strategies. |
PDF Full Text Request