Biclustering gene expression data by random projection based on bucketing

It is an important task for biologists to analyze gene expression data with microarray technology development. Biclustering of gene expression data is the process of grouping a subset of genes over a subset of conditions into a class, in which each gene behaviors similarly over the selected conditions and each condition is related to a certain classification. In this paper, we present a random projection method to find the largest biclusters from gene expression data. To avoid sampling the column uniformly, we adopt the bucketing technology to estimate the probability to sample each column. Experiments show that our method can find the largest biclusters in simulation data and real data.