Combine Bi-clustering Method Of Gene Expression Data

With the development of high throughput detection technology, such as DNA microarray and the nucleotide chip, people can quantitatively detect the production of gene transcripts(mRNA) from the whole genome level, leading an explosive growth of gene expression data. The problem that how to do effective analysis, dig out useful information from the gene expression data become the focus of research in post genomic era. One main task of gene expression data analysis is classify the genes according its gene expression level(or classify the experimental conditions) to obtain the genes(or experimental conditions) which has biological significance. Based on the massive gene expression data which produced by gene chip technology,clustering analysis, which emerge as the times require, was widely applied to solve the above problem. As a new branch of clustering analysis, bi-clustering has opened a new perspective for us. Bi-clustering analysis is a hot problem in data mining, and has important application in gene expression data analysis[3,6]. Bi-clustering analysis is to find sub matrix from data matrix, such that there is a certain consistency between elements of the sub matrix. Bi-clustering analysis an important means of gene expression data processing, however,bi-clustering analysis must be simultaneous clustering the rows and columns,which has proved that it is a NP( non-deterministic polynomial)-hard problem[7].Except for the CC algorithm[6], the mature algorithm is rarely. This paper present an algorithm which based on combined two-way clustering model, the method identify a group of genes which has the same change trend under different conditions, through simultaneous clustering the genes and conditions.This paper test the performance of the method(use human brain samples’ data), and compare the experimental results with several classic algorithm.The results show that: compared with the classic methods, as time consumption does not increase too much, the quality of clustering results have a large increase, the expression pattern is better, and have more value for the follow-up study.