Detecting coherent local patterns from time series gene expression data by a temporal biclustering method

Time-series gene expression data analysis plays an important role in bioinformatics. In this paper, we propose a biclustering method to detect local expression patterns in time-series gene expression data by performing clustering on both gene and time dimensions. Our method aims to find gene subsets which show coherent expression profiles in some time subsets which have a consecutive order in a bicluster. Specifically, our temporal biclustering method is composed of a discretization procedure and a follow-up sequence alignment, which can identify similar local expression profiles and further reveal coherent local relations such as complementary and time-lagged coherence. We apply our method to yeast cell cycle data, and find several biologically important biclusters.