Research On The Difference Analysis Method Of The Microarray Data In Disease

Disease has threatened the health of humans for a long time. Cell comprises im-portant biological molecules. Therefore, research on disease in molecular level could reveal the essence of life activities. As the important regulatory mechanism of body growing, gene expression level can reflect the current state of cell. With the accelerat-ing development of microarray technology, there have been a large number of human gene expression data in different phenotypic state. Therefore, expression data from dis-ease versus normal samples can help researchers to understand disease mechanisms and detect disease therapy targets. Moreover, expression data from samples before versus after drug treatment may facilitate the prediction of treatment outcome.The human body in disease state can cause abnormal cell activities, including sig-nificantly differential expression of key genes, or disorder state of the metabolic and signal molecules enzymatic reaction. Therefore, in order to research on disease, we need to do systematic analysis from two biological molecule levels, one is the single gene level, and the other is pathway level which performs activity function. Based on gene expression data from different disease-related phenotypes, this paper carried out a number of studies in difference analysis methods, such as signaling pathway analy-sis, significantly differentially expressed gene analysis and application of classic chip processing methods. The main works include:(1) According to the weakness that most existing analysis methods ignored the pathways’overall state, this study proposed a new signaling pathway analysis model based on information divergence, and evalu-ated the proposed approach on nine separate cancer expression datasets. Compared with conventional methods, experiments validated that the new proposed method had higher reproducibility, better specificity, sensitivity and applicability. (2) The regula-tory activity calculated by gene expression data may be different from one obtained by biological experiments. In order to eliminate such inconsistent problem, we pro-posed a new signaling pathway analysis method based on random walk, and applied it on four colorectal cancer gene expression datasets. The experiments indicated the reasonability of using random walk method to quantify real regulatory activity. More-over, compared with state-of-the-art methods, this method had a higher accuracy rate, sensitivity and specificity. (3) With regard to the study on detecting molecule therapy targets, we adopted traditional methods, such as significantly differentially expressed gene analysis and gene set enrichment analysis methods, to detect the association and difference between six colorectal cancer related gene expression data. The study has discovered that there was a great difference in anti-cancer drug treatment outcomes be- tween primary cancer cells and mctastatic cancer cells. and has found out the crucially differentially expressed genes and significantly disturbed activity units caused by drug treatment. The molecules whose state was significantly differential could be predicted for molecular therapy targets.