A Way to Apply Traditional Clustering Methods in Bi-Cluster Detection

Bi-clustering microarry data is very important for the research on gene regulatory mechanisms. Genes which exhibit similar patterns are often functionally related. Compared with the traditional data clustering methods, bicluster detection is highly different. The former usually use the whole row or column vectors as feature vectors. While as to the latter the elements in a founded bicluster are greatly distributed among the original data matrix. Experiments have shown that when the number of the biclusters is unknown even an obvious bicluster which can be found by human beings may be wrongly clustered. In this paper a novel bicluster detection framework is proposed, which tries to make fully use of the rich and powerful existing data clustering methods. The clustering results are used for further bicluster detection. Based on an interesting phenomenon which can be proved to be a characteristic of bicluster, the biclusters are detected one by one. At the end of the paper experiment results on the simulated data are presented.