Randomized Algorithmic Approach for Biclustering of Gene Expression Data

Microarray data processing revolves around the pivotal issue of locating genes altering their expression in response to pathogens, other organisms or other multiple environmental conditions resulted out of a comparison between infected and uninfected cells or tissues. To have a comprehensive analysis of the corollaries of certain treatments, deseases and developmental stages embodied as a data matrix on gene expression data is possible through simultaneous observation and monitoring of the expression levels of multiple genes. Clustering is the mechanism of grouping genes into clusters based on different parameters. Clustering is the process of grouping genes into clusters either considering row at a time( row clustering) or considering column at a time(column clustering). The application of clustering approach is crippled by conditions which are unrelated to genes. To get better of these problems a unique form of clustering technique has evolved which offers simultaneous clustering (both rows and columns) which is known as biclustering. A bicluster is deemed to be a sub matrix consisting data values. A bicluster is resulted out of the removal of some of the rows as well as some of the columns of given data matrix in such a fashion that each row of what is left reads the same string. A fast, simple and efficient randomized algorithm is explored in this paper, which discovers the largest bicluster by random projections.