Application Of Complex Network Analytical Method To Genome-Wide Association Studies

Genome-wide association studies have played very important roles in identifying gene mutations associated with complex human diseases in the whole genome level.In recent years,a large number of algorithms about genetic locus detection have emerged in this field.Although these algorithms have achieved great success,there are still some problems.On the one hand,various approaches perform significant differently on different disease models.One of the important reasons is that the existing approaches were constructed on only one correlation model while the different disease models vary a lot.Thus these methods result in low power and a high false-positive rate.On the other hand,Traditional case-control studies mainly analyze the correlation between individual single nucleotide polymorphism(SNP)and disease,and ignore the complex interactions between the SNPs.To solve the aforementioned problems,the thesis presents a key SNPs selecting algorithm based on mutual information.In this thesis,we constructed reversely the SNPs interaction network using simulation data based on the theory of mutual information and compared the difference of the statistics of SNPs interaction networks between case and control groups with the increase of the mutual information threshold.For a specified threshold,the SNPs,which made significant contribution to the network structure,were selected to form the set of "structural key SNPs" of the SNPs interaction network.Under the premise of pre-set disease SNPs,we simulated large amounts of case-control data by running the HAPGEN2 using the information of BRCA2 gene on chromosome 13 from HapMap.Based on the mutual information between two SNPs,we respectively constructed the SNP interaction networks for case and control groups using the simulated data.The comparison of the six network statistics showed that,within the considerable range of the mutual information threshold,modularity,average path length,clustering coefficient and average of vertex betweenness of the case are significantly different from those of control as the threshold increased.For the interaction network with a given threshold,we computed the degree of each SNP in case and control networks and then did a subtraction.Set a value of,we filtered the SNPs to form the "structural key SNPs" set.The results of experiments with different showed that the chosen value could select the key SNPs having large influence on network structure and the method could efficiently select the pre-set disease SNPs.At the end,we used genes of brain glioma to illustrate the effectiveness of the method and detected some pathogenic genes,such as NF2 and TP53,and give some prediction.