Study On Comparison Of Bayesian Network Meta-analysis Methods For Polygene Genetic Association Studies

Objective: Network meta-analysis is a methodology developed on the basis of traditional meta-analysis in recent years.Traditional meta-analysis can only analyze and compare two treatment factors.Network meta-analysis can compare multiple treatment factors and filter the best treatment factors based on the ranking results.At present,the network meta-analysis of randomized controlled trials is more comprehensively applied and developed,which has a variety of analysis methods based on both frequency and Bayesian.However,due to the characteristics of genotype data in genetic association studies,the current methodological research on the network meta-analysis of genetic association studies has to be further developed.The development of the network meta-analysis method of genetic association research can provide a new method of evidence integration and make genotype data more effective used,so as to explore the role of genetic factors in the occurrence and development of diseases,and to provide theoretical and application basis on disease diagnosis,prognosis and treatment.Methods: According to the characteristics of genotype data,this study proposed a ranking comparison method,a combined effect size method and an ANOVA model method based on the Bayesian.According to the three methods,the network meta-analysis of polygene genetic association research was realized.Through literature search and data extraction to obtain relevant genotype data,according to the NOS scale to evaluate the quality of literature that meet the standards.Two selected examples of the correlation study on the micro RNA gene polymorphisms and liver cancer susceptibility,and DNA repair gene polymorphisms and the efficacy of platinum-based chemotherapy for non-small cells lung cancer were analyzed and verified.For the statistical analysis,first,traditional meta-analysis was performed,and then a network plot describing the association of gene polymorphisms was drawn.According to the proposed ranking comparison method,the combined effect size method and the ANOVA model method,the corresponding models were constructed for Bayesian network meta-analysis,the effect indicators were calculated to evaluate the relationship between gene polymorphism and disease.The trace plot,density plot,and Brooks-Gelman-Rubin diagnosis plot were used to evaluate model convergence and fitting effects.Statistical software R and STATA13.0were used.Results: In the study on the association between micro RNA gene polymorphisms and HCC susceptibility,20 studies were screened,including four micro RNA gene polymorphisms mi R-146a(rs2910164),mi R-149(rs2292832),mi R-196a2(rs11614913),mi R-499(rs3746444).In traditional meta-analysis,only mi R-196a2 rs11614913 is associated with hepatocellular carcinoma susceptibility(mi R-196a2 rs11614913:TC +CC vs.TT : OR = 1.232,95%CI = 1.028-1.476).The results of the ranking and comparison methods respectively suggested that mi R-149 rs2292832 or mi R-196a2rs11614913 were genetic polymorphisms that might be related to the occurrence of hepatocellular carcinoma.The combined effect size method showed that mi R-196a2rs11614913 and mi R-499 rs3746444 were genes related to hepatocellular carcinoma susceptibility(mi R-196a2 rs11614913:TC + CC vs.TT:OR = 1.40,95%CI = 1.10-1.80;mi R-499 rs3746444:TC+CC vs.TT:OR = 1.50,95%CI = 1.10-1.90).In the ANOVA model method,mi R-196a2 rs11614913 were the only gene polymorphism related to hepatocellular carcinoma susceptibility among the four gene polymorphisms(mi R-196a2rs11614913:TC + CC vs.TT:OR = 1.20,95%CI = 1.01-1.43).In the study on the association between DNA repair gene polymorphisms and the efficacy of platinum-based chemotherapy for non-small cell lung cancer,26 studies were screened,including seven DNA repair gene polymorphisms ERCC1(rs11615),XPD(rs13181),XRCC1(rs1799782),ERCC2(rs1799793),XRCC1(rs25487),ERCC1(rs3212986),XRCC3(rs861539).Traditional meta-analysis showed that the three gene polymorphisms XRCC1(rs1799782),XPD(rs13181)and XRCC3(rs861539)gene polymorphisms were related to the curative effect of non-small cell lung cancer,and the other four have no significant difference(XRCC1 rs1799782:CT+TT vs.CC:OR = 1.391,95%CI = 1.134-1.705;XRCC1 rs25487 : GA+AA vs.GG : OR = 1.589,95%CI =1.331-1.896;XRCC3rs861539:CT+TT vs.CC:OR = 1.407,95%CI =1.067-1.856).The ranking comparison method showed that the probability ranking of each gene polymorphism changed,however it is difficult to determine which gene polymorphisms were related to the therapeutic effect of non-small cell lung cancer chemotherapy.The combined effect size method found that XPD(rs13181),XRCC1(rs1799782),XRCC1(rs25487)were related to the efficacy of platinum-based chemotherapy for non-small cell lung cancer(XPD rs13181:AC+CC vs.AA:OR = 0.86,95%CI = 0.74-0.99;XRCC1 rs1799782:CT+TT vs.CC:OR = 1.30,95%CI =0.74-0.99;XRCC1 rs25487:GA+AA vs.GG:OR = 1.50,95%CI =1.30-1.70).In the ANOVA model method,XRCC1(rs25487)was the only relevant gene polymorphism(XRCC1 rs25487:GA+AA vs.GG:OR = 1.45,95%CI=1.17-1.79).Conclusion: The ranking comparison method,the combined effect size method and the ANOVA model method can all be used to realize the Bayesian network meta-analysis for genetic association studies,and the interpretation of the results is more adequate than traditional meta-analysis.The network meta-analysis of the ranking comparison method has limitations in which the combined effect size cannot be.The network meta-analysis of the combined effect size method can obtain the combined effect size OR,but the original data information is lost.The network meta-analysis of the ANOVA model method uses all the original data of the study to integrate direct and indirect evidence could obtain the combined effect size index and better explain the association between genetic polymorphism and disease based on the superiority index.Among the three methods,ANOVA model method is most suitable for the network meta-analysis of genetic association studies.