Recently, several methods have been suggested to discover biclusters from gene expression data matrices, where a bicluster is defined as a subset of genes that exhibit a highly correlated expression pattern over a subset of conditions. Most of them produce sub-optimal biclusters with greedy or stochastic approach. In contrast, we propose a new biclustering method, BiModule, that exhaustively searches biclusters in a realistic time based on a closed itemset mining algorithm. Comparative experiments to salient biclustering methods are performed to test the validity of biclusters extracted by BiModule using synthetic data and real expression data. We show that BiModule provides high performance compared to the other methods in extracting artificially-embedded modules as well as modules strongly related to GO annotations and protein-protein interactions.