DYNAMIC CORE BASED CLUSTERING OF GENE EXPRESSION DATA

Modern microarray technology allows measuring the expression levels of thousands of genes, under different environmental conditions and over time. Clustering is, often, a first step in the analysis of the huge amounts of biological data obtained from these microarray based experiments. As most biological processes are dynamic and biological experiments are conducted during longer periods of time, the data is continuously subject to change and researchers must either wait until the end of the experiments to have all the necessary information, or analyze the data gradually, as the experiment progresses. If the available data is clustered progressively, using clustering algorithms, as soon as new data emerges, the algorithm must be run from scratch, thus leading to delayed results. In this paper, we approach the problem of dynamic gene expression data sets and we propose a dynamic core based clustering algorithm, which can handle newly collected data, by starting from a previously obtained partition, without the need to rerun the algorithm from the beginning. The experimental evaluation is performed on a real -life gene expression data set and the algorithm has proven to perform well in terms of a series of evaluation measures.