A Hybrid Possibilistic Algorithm for Biclustering: Application to Microarray Data Analysis

A attractive way to perform biclustering of genes and conditions is to adopt the notion of fuzzy sets, which is useful for discovering overlapping biclusters. Fuzzy clustering is well known as a robust and efficient way to reduce computation cost to obtain the better results. However, this approach is not explored very well. In this paper, we propose a new algorithm called, Refine Bicluster for biclustering of microarray data using the fuzzy approach. This algorithm adopts the strategy of one bicluster at a time, assigning to each data matrix element, i.e. each gene and for each condition, a membership to bicluster. The biclustering problem, in where one would maximize the size of the bicluster and minimize the residual, is faced as the optimization of a proper functional. Applied on continuous synthetic datasets, our algorithm outperforms other biclustering algorithms for microarray data.