Construction And Analysis Of Complex Networks For Genome-wide Association Data

Diseases have seriously affected human health since the emergence of human beings.Genome-wide association study(GWAS)can locate the single nucleotide polymorphisms(SNPs)and other genetic markers associated with complex diseases,which is helpful to further understand the pathogenesis of diseases and provide great help for prevention,diagnosis and therapy of diseases.So it is a hot spot in biomedical research nowadays.The human genome is a very large system from the perspective of SNPs.There are complex relationships among SNPs,but a large number of Genome-wide association studies only focus on the association between one single SNP and complex diseases,and ignore the complex interactions between SNPs.Therefore,we carried out genome-wide association study based on conditional mutual information method(CMI)and complex network analysis method,,which could fully utilize the complex interactions between SNPs,and eventually select the possible causal SNPs set with a high accuracy rate,which is of great significance for locating SNPs related complex disease in genome-wide association data.In this thesis,we first pre-processed SNP data of gene BRCA2 and ASAH1 related to breast cancer and schizophrenia respectively.Then we simulated case group and control group according to the pre-processed SNP data and constructed two groups of SNP-SNP interaction networks according to the conditional mutual information matrices of case group and control group.Next we compared the two groups of networks using network statistics,and determined the optimal CMI threshold which can effectively distinguish the two groups of networks.And we also located the possible causal SNPs through the centrality difference of nodes between the two groups of networks.Then we evaluated the accuracy of our method under different simulation schemes and data scales,and compared our method with other methods based on the data of gene BRCA2.The results showed that the possible causal SNPs sets obtained by our method are small and contain the preset causal SNPs,and the accuracy of our method is superior to that of the parameter selection method based on mutual information(MI).Finally,we summarized the main work and innovation of this thesis,and prospect the future work according to the deficiency of our research.