BiETech : Bicluster Ensemble Techniques

Abstract

Various biclustering algorithms have emerged now a days that try to deliver good biclusters from gene expression data which satisfy a particular objective function. Users are lost in finding the best out of these algorithms. Ensemble techniques come to rescue     of these users by aggregating all the solutions and providing a single solution which is more robust and stable than its constituent solutions.  In this paper, we present two different ensemble techniques for biclustering solutions. We have used classifiers in one approach and the other approach uses the concept of metaclustering for forming the consensus. Experiments in this research are performed   on synthetic and real gene expression datasets as biologists are interested in finding meaningful patterns in expression of genes.  The experiments show that both the approaches proposed in the paper show improvement over the input solutions as well as the existing bicluster ensemble techniques.