Characterizing gene coexpression modules in Oryza sativa based on a graph-clustering approach

Recent advances in genome research have yielded a vast amount of large-scale data (e. g. DNA microarray) and have begun to deepen our understanding of plant cellular systems. Meta-analysis such as gene coexpression across publicly available microarrays has demonstrated that this approach is useful for investigating transcriptome organization and for predicting unknown gene functions in biological processes ranging from yeast to humans. However, no overall coexpression-network module in rice has been examined in detail. Here we present the coexpression clusters of rice genes based on unbiased graph clustering of the coexpression network of 4,495 genes. The coexpression network was constructed by using over 230 microarrays; it manifested several properties of a typical complex network (e. g. scale-free degree distribution). Using the DPClus algorithm that can extract densely connected clusters we detected 1,220 clusters. We evaluated these clusters using gene ontology enrichment analysis. We conclude that this approach is important for generating experimentally testable hypotheses for uncharacterized gene functions in rice and we posit that meta-analysis across publicly available microarrays will become increasingly important in crop science.