Analysis Of Arabidopsis Gene-Related Network Under Stimulus

It is important to understand the physiological mechanism of genome response to external stimulus by studying the change of Arabidopsis genome expression under stimulus. In this dissertation, by the method of reverse network modeling, the mutual information networks of normal (control group) and stimulus (experimental groups) of Arabidopsis abstract genes and flowering-related genes are constructed respectively. Then analyze and study the structural properties of the networks. The details are as follows:For the mutual information networks of Arabidopsis abstract genes, through statistical analysis and comparison, the significant difference is found between the network structures of control group and experimental groups, and statistics are given which can distinguish the two types of networks:average degree, clustering coefficient, modularity, the proportion of non-isolated node. What's more, the method to classify the training sets of normal and each stimuli in multi-parameter space is proposed. The existence of differences is confirmed from the node of view of the whole network structure.For mutual information network of Arabidopsis flowering-related genes, through statistical analysis and comparison, the significant difference of the network structures between control group and experimental groups is found, and the statistics that can distinguish the two types of networks are given:average degree, average coreness. Then, the method to mine "the structural key genes" of the flowering is given. Using the method, the patterns of coreness and degree "the structural key genes" are obtained which contribution the two types of patterns the largest. The numbers of key genes are 8 and 11 respectively. Furthermore, the 8 genes are fully included in the 11 genes. That structure determines function inspirits the gene whose pattern is the same to the pattern of database may play positive role on flowering and the gene whose pattern is contrary to the pattern of database may play negative role on flowering. The information from TAIR database indicates that there are 8 genes out of 11 genes satisfying the law. That is, the efficiency to predict of the method is 72.73%. Therefore, the method can predict the functions of the remaining genes. In the research, the law is found:when the coreness and degree of gene in the long day experiment is higher than normal condition, the genes play a positive role on flowering; conversely, when the coreness and degree of gene in the long day experiment is lower than normal condition, the gene inhibits the flowering. The information from TAIR database indicates that there are 5 genes out of 8 genes satisfying the law. That is, the efficiency to predict of the method is 62.5%. Thus, the method can predict the functions of genes, too. When the methods of network modeling and model analysis are used for analyzing other plants' expression profile data under stimulus, it is universal significant to understand the mechanism of genome under external stimulus.