An EA framework for biclustering of gene expression data

In recent years, several biclustering methods have been suggested to identify local patterns in gene expression data. Most of these algorithms represent greedy strategies that are heuristic in nature: an approximate solutions is found within reasonable time bounds. The quality of biclustering, though, is often considered more important than the computation time required to generate it. Therefore, this paper addresses the question whether additional run-time resources can be exploited in order to improve the outcome of the aforementioned greedy algorithms. To this end, we propose a general framework that embed such biclustering methods as local search procedures in an evolutionary algorithm. We demonstrate on one prominent example that this approach achieves significant improvements in the quality of the biclusters when compared to the application of the greedy strategy alone.