Measuring The Similarity Of Co-regulated Genes By Integrating Quantity And Tendency Of Gene Expression Changing

With the development of Microarray technology, more and more gene expression date are generated for human to research. A question followed is how to analyses the date efficiently and obtain the information inside the biologic microcosmic world with the purpose to master the rules of living activity which could be useful. It is popular reconstructing the gene regulatory network by using gene expression date in bioinformatics. To do so, thus, we can penetrate into finding and predicting the gene function. Further more, it could descript the true process in living activity.The work of reconstructing the gene regulatory network involving two important elements, the one is defining the relation among genes, and the other one is constructing the network model. The first one which including the date pre-processing and the clustering analysis is the key step for the next. Date pre-processing works according to a group of standard steps, and generates a group of normalized date vectors, called gene expression profile. Thereby, clustering analysis can works on it to digging information, defining the relation among genes, and finding co-regulated genes primarily, by which we can do the next other works.However, the key factor about how to find out the co-regulated genes would seriously affect the results of clustering analysis, which was found in the past researching. Such problem would directly decrease the authenticity and the value of the gene regulatory network being constructed because of missing the true relationship among genes. Thus, this paper would focus on finding out more co-regulated genes.The traditional research based on the hypothesis that the co-expressed genes (the genes with similar expression value in most watching time point) are the co-regulated genes (the genes regulated by at least one transcription factor), but it is not absolute truth. Some recently research illuminates that the two kinds of genes are not equal, while the various factors lead to the co-expression during the complex transcription regulatory process, therefore, neither the co-expressed genes should be regard as the co-regulated genes simply, nor the co-regulated genes should be co-expressed too. There exist positive co-regulated genes and negative co-regulated genes in real co-regulated genes: positive co-regulated genes are always increasing or decreasing with expression value in the same time, called positive consistency; negative co-regulated genes are always changing with contrary direction, we called it negative consistency. However, co-regulated genes are found expressed in different expression level which could not be considered the co-repressed genes, showing us the hypothesis has its limitation again.According to the feature of co-regulated genes in expressing, the traditional methods which measuring the similarity of genes'expression date only by adjusting the similarity of numerical value fail to integrate all the co-regulated genes mentioned above. Though, Pearson Correlation Coefficient can partially find these co-regulated genes, the experiment results by using it still have some shortage, while gene expression date as a group of correlated date having time serial information, ignoring it will results in the information missing. Some research focus on the time serial feature in the genes expression date in recent makes a advancement in the work but ignore the quantity in genes'expression changing.The quantity and the tendency are both integrated in this paper, and the consistency in changing is scored to measure the similarity of genes, by which we defined the correlation coefficient SMBS. According to the SMBS, the more consistency changing among genes the higher correlation coefficient they would have. Obviously, the positive and negative co-regulated genes should be recognized the co-regulated genes, so as to difference expressed co-regulated genes. So, more co-regulated genes will be found out using SMBS. We applied the SMBS in K-means cluster algorithm using real Saccharomyces cerevisiae expression date, the experiment result has verified in SCPD transcription factor database, showing that all kinds of co-regulated genes are found and meeting the real regulatory mechanism in yeast gene. The result compared with the popular clustering analysis software CLUSTER3.0 in the same experiment condition, showing the fewer running time and higher veracity.We can see, integrating the quantity and tendency in the genes'expression changing adequately dig the abundant information in the gene expression date, while providing the brand-new conception and horizon to measuring the similarity of gene expression date. Being used in widely, the method would enhance the authenticity and reliability rapidly, more importantly, it would be helpful to master the rules in gene transcription process and predict the gene function, and then firmly support the explore the further research.The organizational structure of this paper is as follows:Chapterâ… : Introduction. This chapter synoptically introduced the knowledge background of gene regulatory network and the works in this paper.Chapterâ…¡: The gene expression profile. This chapter described the experiments of Microarray, the form of gene expression date, and the Saccharomyces cerevisiae gene expression date.Chapterâ…¢: The clustering analysis of gene expression date. This chapter described the correlation similarity measurement and the clustering algorithm in the clustering analysis, showing how to analyses the gene expression using clustering methods.Chapterâ…£: The clustering analysis basing on the time serial feature in gene expression date. This chapter defined the correlation coefficient SMBS of gene expression date, and then applied it to the real date.Chapterâ…¤: Summary and Outlook. Summarizes the main research work of this paper, and expectation the future works.