cHawk: An Efficient Biclustering Algorithm based on Bipartite Graph Crossing Minimization

Biclustering is a very useful data mining technique for gene expression analysis and profiling. It helps identify patterns where different genes are co-related based on a subset of conditions. Bipartite Spectral partitioning is a powerful technique to achieve biclustering but its computation complexity is prohibitive for applications dealing with large input data. We provide a connection between spectral partitioning and crossing minimization which is amenable to efficient implementations. Theoretical construction of Biclustering model based on crossing minimization is provided. Based on this model, an efficient biclustering algorithm, which is termed as cHawk, is developed. We have evaluated cHawk on both synthetic and real data sets. We show that cHawk is able to identify, with good accuracy, constant, coherent and overlapped biclusters amid noise. Moreover, its execution time grows linearly with input data size.