Construction And Cluster Analysis Of Gene Regulatory Network

This thesis researches the construction and cluster analysis of gene regulatory networks based on gene chips data. By reliable construction method, we find the gene expression which display the interrelationship between genes best, then we do comparative analysis to get the structure and properties of gene regulatory networks under different condition, thus we work out the biological meaning of the GRNs. We proposed a new method to construct the gene regulatory network, which decrease the computation complexity and increase the reliability, which is count the p-value with checking by permutation methok.The reliability of the construction method will affect the analysis results of following research. Out research employ a strict testing by permutation and a reliable checking p-value by permutation of p-value, and we use Kuiper test to show the rationality of constructing network with testing by permutation. With the assumption of Hypothesis testing, we choose the Pearson coefficient as the measure of correlation, select adequate sampling to compute the p value, and then construct the network with PCER. We show that our result is as good as the result of total replacement.After computing the p-value, we define an threshold value T, and add an edge to the network if the p-value of two genes is less than T. With genes as the nodes, and the relationship between genes as the edges, we successfully construct the gene regulatory network.On the one hand, we do clustering analysis with Newman Fast Algorithm on the constructed networks. Newman Fast Algorithm is a greedy approach, which doesnt promise to give out the best results, so the clustering analysis is with noise. Since those clusters with few genes are vulnerable to noise and leads to conclusion with less believe, our research focus on those cluster with more genes. On the other hand, we compare the clustering coefficient and the number of network motif to get the structure difference of the gene regulatory network under different condition, thus analyze its biological meaning.Out research analyzed the leukocyte data of acute myeloid leukemia and normal condition, and draw the conclusion that the GRN of acute myeloid leukemia is much more random and discrete than normal condition, which signifies that the GRN of acute myeloid leukemia regulatory networks is lack of some mechanism, which cause cancer.The correctness of our proposed methods is shown since our conclusion is in accord with the basic principles of biology.