Research For Auxiliary Diagnosis Of Hepatocellular Carcinoma Patients Based On KIF1BmRNA、STAT4mRNA And Gene Expression Programming

Objective: To investigate the expression and clinical significance of kinesin family member 1B(KIF1B) m RNA and signal transducers and activators of transcription(STAT4)m RNA in the peripheral blood of patients with hepatocellular carcinoma, cirrhosis groups,chronic hepatitis patients and normal groups, and then evaluated the application value of KIF1 B m RNA and STAT4 m RNA in diagnosis of hepatocellular carcinoma(HCC).Additionally, common serum tumor markers AFP, carcinoembryonic antigen(CEA) and carbohydrate antigen 199(CA199) and biochemical indexes such as liver function indicators were used gene expression programming(GEP) to establish a prediction model of HCC, which was performed to the auxiliary diagnosis for HCC.Methods: 1. A total of 352 peripheral blood samples were collected, including 126 cases of HCC, 80 cases of liver cirrhosis, 46 cases of chronic hepatitis patients, 100 cases of healthy controls, from Sep. 2013 to Jul. 2014 in Affiliated Hospital of Qingdao University Medical College. Reverse Transcription-Polymerase Chain Reaction(RT-PCR) was used to detect the peripheral blood level of KIF1 B and STAT4 m RNA in 352 peripheral blood samples. SPSS 17.0 software was applied to analyze KIF1 B and STAT4 m RNA difference in the expression of each group, and draw the receiver operating characteristic(ROC)curve to determine the optimal cut-off value of KIF1 B, STAT4 and AFP in diagnosing HCC and the sensitivity and specificity when combination of three biomarkers.2. A total of 1000 tumor markers and biochemical indicators were collected,including 500 cases of HCC and 500 cases of healthy controls, from Jan. 2011 to Jul. 2014 in Affiliated Hospital of Qingdao University Medical College. GEP algorithm was used toAPS 3.0 software, and established the model of early diagnosis for HCC. And the reliability of GEP algorithm was also introduced artificial neural network(ANN) analysis method.Results: 1. In HCC group, the peripheral blood of KIF1 Bm RNA expression level was lower in the serum higher CA199 level group(P<0.05). However, the peripheral blood level of KIF1 Bm RNA was uncorrelated with age, gender, tumor TNM stage, tumor size,whether the HBs Ag positive, lymph node metastasis, tumor thrombus of the portal vein, serum AFP and CEA level(P all>0.05). The peripheral blood level of STAT4 m RNA was higher in the older patients(P < 0.05). However, the peripheral blood level of STAT4 m RNA was uncorrelated with gender, tumor TNM stage, tumor size, whether the HBs Ag positive, lymph node metastasis, tumor thrombus of the portal vein, serum level of AFP, CEA and CA199(P all>0.05).2. The KIF1 Bm RNA level in HCC group was significantly lower than liver cirrhosis group, chronic hepatitis group and healthy control group(F=22.27, 17.73, 20.40; P all<0.05). There was no significant statistical difference of KIF1 Bm RNA level among liver cirrhosis group, chronic hepatitis group and healthy control group(P>0.05). It was no significant statistical difference of STAT4 m RNA between HCC group and healthy control group. However, the level of STAT4 m RNA in HCC group compared with liver cirrhosis group and chronic hepatitis group, the result was no statistical difference. There was also no significant statistical difference of STAT4 m RNA level among liver cirrhosis group,chronic hepatitis group and healthy control group(P>0.05).3. The area under the ROC curve of KIF1 Bm RNA, STAT4 m RNA and AFP were0.706, 0.536, 0.838 respectively. The area under the ROC curve of STAT4 was 0.536,which did not used to the diagnosis of disease. As the area under the ROC curve of combination of KIF1 B and AFP was 0.875. However, there was no significant difference of the area under the ROC curve between AFP alone and KIF1 B combined with AFP(P>0.05).4. The GEP model 1 incorporating six biomarkers was established by using APS 3.0software, including AFP、CEA、CA199、ALB、ALT and Cl-. The accuracy of the GEP model 1 training set was 95.75%, the accuracy of the GEP model 1 test set was 92.5%.The GEP model 2 incorporated four biomarkers, including AFP、CA199、ALB、ALT. The accuracy of the GEP model 2 training set was 95.75%, the accuracy of the GEP model 2test set was 92.5%. In the method of ANN, model 1 also incorporated six biomarkers,including AFP、CEA、CA199、ALB、ALT and Cl-. The training and test set accuracy of Multi-layer Perceptron(MLP) in model 1 was 91.6% and 89.1%, respectively. The training and test set accuracy of Radical Basis Function(RBF) in model 1 was 91% and83.3%, respectively. The model 2 incorporated four biomarkers, including AFP、CA199、ALB、ALT. The training and test set accuracy of MLP in model 2 was 90.5% and 85.4%,respectively. The training and test set accuracy of RBF in model 2 was 90% and 77.8%,respectively.Conclusion: 1. The peripheral blood level of KIF1 Bm RNA was lower expressed in the patients with HCC, and the level of KIF1 Bm RNA was lower in serum higher level of CA199.2. There was no significant statistical difference of KIF1 Bm RNA level among HCC group, liver cirrhosis group, chronic hepatitis group and healthy control group.However, the STAT4 m RNA level was higher in the older age of HCC patients.3. The diagnosis accuracy on HCC of peripheral blood level of KIF1 Bm RNA and STAT4 m RNA were less than serum AFP. Therefore, the peripheral blood level of KIF1 Bm RNA and STAT4 m RNA were not as the potential biomarkers for the diagnosis of HCC.4. The model which was established by GEP method in combination with six serological indexes for the early diagnosis of liver cancer has higher accuracy.5. GEP model for auxiliary diagnosis of HCC was superior to the ANN method, it was more suitable for analysis of clinical data. It can be used to the early auxiliary diagnosis of HCC.