Higher order structure in the cancer transcriptome and systems medicine

Abstract

Mol Syst Biol. 3: 94 A formal goal of gene expression studies in cancer is to reconstruct the pathways that characterize tumor behavior through its transcriptional signature (Liu, 2005; Liu et al , 2006). The ultimate utility is not only to discover molecular components of these cancers, but also to exploit this knowledge to improve therapeutics (Bild et al , 2006a). To accomplish this, a number of approaches have been used to achieve higher level order in the expression of gene sets (Mootha et al , 2003; Segal et al , 2004; Subramanian et al , 2005; Bild et al , 2006b). Recently, Tomlins et al (2007) made a further contribution to this effort by devising an approach to construct a molecular concept map (MCM) in the analysis of human prostate cancers. In this work, they microdissected 101 prostate cancers that ranged from benign epithelium to metastatic disease and assessed the expression of transcripts on a genome‐wide scale using printed cDNA microarrays. Approximately 14 000 molecular ‘concepts’ representing biologically connected genes were used to analyze the expression profiles and identify a number of relevant pathways that might drive prostate cancer biology. As in many good scientific articles, several layers of importance that are significant to different research communities can be found. Pertinent to this journal, this paper by Tomlins et al (2007) provides an instructive focal point for the discussion of fundamental problems that bedeviled systems biologists working on primary human cancers. For example, given the complexity of mammalian systems and the further genetic scrambling that takes place with cancer, is it possible to develop systems approaches that can solidly map, in cancer cells, interactions relevant to clinical cancer biology? Does MCM provide definitive answers to these problems? The honest answer must be no, but …