Computational methods for discovering correlated gene expression patterns with parallelized individual dimension-based clustering of microarray data

Abstract

Microarray analysis is used to monitor expression patterns of tens of thousands genes simultaneously. It is a significant process to identify correlated gene expression patterns in microarray analyses as it helps to reveal novel function of genes, gene expression regulation and concerted gene functions in pathogenesis. Although large microarray datasets have recently ap-peared to be common, existing methods are not able to process large microarray datasets due to their considerable computational complexity and memory requirements. Furthermore, typical clustering methods construct oversimplified clusters that ignore subtle but meaningful changes in the expression patterns in large microarray datasets. Thus, in order to examine extensive micro-array datasets, it is required to develop an efficient clustering method that is able to identify not only absolute expression differences but also expression profile patterns at different expression levels. Moreover, a number of biclustering algorithms have been developed to search biclusters that have similar gene expression patterns in a subset of conditions. However, limitations are found as correlated gene expression patterns cannot be highlighted by those algorithms; they merely focus on finding similar gene expression levels. Although a few correlation-based biclustering algorithms have been proposed, they are able to extract biclusters in a limited search space and produce uneven biclustering results. In this thesis, thus, we propose two significant gene expression pattern mining algorithms that we devised, CLIC, and BICLIC in order to overcome the shortcomings found in conventional algorithms, Firstly, an individual dimension-based clustering method, CLIC, will be presented, which is able to discover globally correlated patterns of microarray data. The most significant advantage of conducting CLIC is that not only is it able to meet particular requirements of clustering analysis but also not limited to large micro...