Global analysis of transcriptional changes in human cancer using the gene coexpression network

Abstract

With recent interest in biological networks, the gene coexpression network has emerged as a novel holistic approach for microarray analysis. The recent availability of large amount of microarray data on human cancer allowed us to collect 10 tumor-normal datasets spanning 13 tissue origins. To explore tissue-wide coexpression pattern, we integrated these datasets by employing meta-analytic methods, which had proved to be reliable and robust against noises of microarray data in previous differential expression studies. We constructed two distinct coexpression networks whose edges connected genes consistently coexpressed over tumor and normal samples respectively. The comparison of the two networks revealed that cancer induces many significant changes in coexpression relationships. We observed increased coexpressions among genes involved in the cell cycle and DNA/RNA/protein synthesis, and decreased coexpressions among genes involved in the signal transduction, mitochondrial energy metabolism, and plasma membrane functions. In addition, it was suggested that the aberrant coexpression of ubiquitin ligases and associated cell-cycle regulators would contribute to tumor development, regardless of cancer type.