Research Of Normalization Strategy Of Expression Data For Gene Chip

Motivation: With the rapid development of molecular biological and computational technologies, the implement of HGP and achievement of HGD show that modern life science research has entered the post-genomic era, researchers focus on from structure genome science to function genome science. The DNA microarray is capable of profiling the expression levels of many genes simultaneously, and is a promising technology for the research of gene interactions.And how to exactly and reasonably analyze the large of data that result from the microarray experiments has become a problem that researchers are eager to resolve, and important research content of bioinformation and main method. Now there have been quite a few methods including sophisticated mathematical models that are used to analyze image and data of gene chip, especially normalization of data analysis earlier, which will be perfected and innovated further. Normalization is the process of removing some sources of variation which affect the measured gene expression levels, it plays an important role in the earlier stage of microarray data analysis, the subsequent analysis results are highly dependent on normalization. Although a number of normalization methods have been proposed, it has been difficult to decide which methods perform best. In our research, therefore, we bring forward some normalization models, emphases on analyzing and compare of normalization strategy from four models, base on its normalization strategy, how to choose method and how to deal with for gene expression chip of different density, finally, set up normalization strategy of expression data for gene chip. Method: Normalization, an important way in data analyzing of gene expression chip, badly affects statistic analysis later, such as clustering. There we give four normalization models: control spots normalization(CSN),total intensity normalization(TIN),locally weighted linear regression normalization(LWLRN) and locally mean normalization(LMN); Analyzing strategy of every normalization model; we get normalization factor through software of image and data that we frequently used, and apply to data for gene expression chip of different density, analyze and compare each method and its results, and search the best normalization. Result: Four normalization methods can effectively reduce influence resulting from systemic errors for gene expression data of different density chip, and make data processed more compare and dependent, by analyzing them, we find: control spots normalization method has very good effect for low density and stable expression chip; total intensity normalization is well used for distributing uniformity of signal intensity of low density chip; locally weighted linear regression normalization can effectively reduces influence of background noise for depend-intensity and high density chip; locally mean normalization can effectively solves the problem of distributing odds of total intensity for high density chip.Conclusion: In our research, it is feasible for these normalization methods to analyze gene expression chip of different density, which can effectively reduce systemic errors and get dependable and credible gene expression level or expression ratio, more exactly find genes of significant difference, and provide important clues for latter research. However, normalization methods which have no uniform standard are being developed and need to be improved further. With the development of non-linear technology,software and hardware of computer, microarray data normalization analysis and transformation will be broken through.