The Research On Biclustering Algorithm Applied To Gene Expression Data

Gene chip (or DNA Microarrys) technologys allow to monitor the gene expression levels of thousands of genes under different biological conditions. Gene expression data obtained form DNA microarry experiments, can reflect cell physiology status , provide massive amounts of heredity message. The general approache of analyzing gene expression data is clustering. Recently, biclustring algorithms have been used to extract useful information form gene expression data. Then biclustring algorithms were applied to drug exploitation, disease cure and early stage diagnose. Along with the develop of DNA microarry technology, the gene expression data sharp increase. How to make use of clustering to obtain conceal information form massive gene expression data? It is the key prolem of biomedicine domain and data mine domain nowadays.The article introduce the basis knowledges of biclustering. Then the article summarized two kinds of opinions which is used to improve biclustering algorithms:matrix factorization and computational intelligence. We conducted the following in-depth study of biclustering algorithm.Firstly, the article improve Interrelated Two Way Clustering by fast nonnegative matrix factorization, put forward a new algorithms called FNMF-ITWC algorithms. The algorithms carry out clustering by fast nonnegative matrix factorization on genes dimensions, reduce dimensions and extract the consistent characteristic of gene expression data which is high dimensions and low samples, Then carry out clustering on samples dimensions, rows and columns iteration find the end of the results.Secondly, the article combined modular singular value decomposition and multi-objective evolutionary algorithms to improve biclustering, put forword a new algorithm called MSVD-MOEB. The algorithm make used of modular singular value decomposition to reducing the dimension of the gene expression data;using multi-objective evolutionary to improve local search strategy; using the extend and merge clustering algorithm to find the end of the result.Finally, the article apply FNMF-ITWC algorithm and MSVD-MOEB algorithm to cancer gene expression data. Through Matlab experiment, discuss the advantage and inadeguate of the two improved algorithms by several validation measures of the result of clustering.