Functional Categories Associated with Clusters of Genes That Are Co-Expressed across the NCI-60 Cancer Cell Lines

et al. (2012) Functional Categories Associated with Clusters of Genes That Are Co- Expressed across the NCI-60 Cancer Cell Lines. PLoS ONE 7(1): e30317. doi:10.1371/journal.pone.0030317 Functional Categories Associated with Clusters of Genes That Are Co-Expressed across the NCI-60 Cancer Cell Lines Barry R. Zeeberg 0 1 William Reinhold 0 1 Rene Snajder 0 1 Gerhard G. Thallinger 0 1 John N. Weinstein 0 1 Kurt W. Kohn 0 1 Yves Pommier 0 1 Ilya Ulasov, University of Chicago, United States of America 0 Current address: Departments of Bioinformatics and Computational Biology and Systems Biology, M.D. Anderson Cancer Center , Houston, Texas , United States of America 1 1 Laboratory of Molecular Pharmacology, Center for Cancer Research, National Cancer Institute, National Institutes of Health (NIH) , Bethesda , Maryland, United States of America, 2 Institute for Genomics and Bioinformatics, Graz University of Technology , Graz , Austria Background: The NCI-60 is a panel of 60 diverse human cancer cell lines used by the U.S. National Cancer Institute to screen compounds for anticancer activity. In the current study, gene expression levels from five platforms were integrated to yield a single composite transcriptome profile. The comprehensive and reliable nature of that dataset allows us to study gene coexpression across cancer cell lines. Methodology/Principal Findings: Hierarchical clustering revealed numerous clusters of genes in which the genes co-vary across the NCI-60. To determine functional categorization associated with each cluster, we used the Gene Ontology (GO) Consortium database and the GoMiner tool. GO maps genes to hierarchically-organized biological process categories. GoMiner can leverage GO to perform ontological analyses of gene expression studies, generating a list of significant functional categories. Conclusions/Significance: GoMiner analysis revealed many clusters of coregulated genes that are associated with functional groupings of GO biological process categories. Notably, those categories arising from coherent co-expression groupings reflect cancer-related themes such as adhesion, cell migration, RNA splicing, immune response and signal transduction. Thus, these clusters demonstrate transcriptional coregulation of functionally-related genes. - Funding: This research was supported by the Intramural Research Program of the National Institutes of Health, National Cancer Institute, Center for Cancer Research, Research and the Austrian Ministry of Science and Research, GEN-AU project Bioinformatics Integration Network. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing Interests: The authors have declared that no competing interests exist. The NCI-60 is a panel of 60 human cancer cell lines that has been used by the Developmental Therapeutics Program (DTP) of the U.S. National Cancer Institute to screen compounds plus natural products since 1990 [1,2]. The NCI-60 panel includes cell lines from colorectal (CO), renal (RE), ovarian (OV), prostate (PR), lung (LC), breast (BR), and central nervous system (CNS) cancer origin, as well as leukemias (LE) and melanomas (ME). We and our many collaborators around the world have profiled the NCI-60 more comprehensively at the DNA, RNA, protein, mutation, functional, and pharmacological levels than any other set of cells in existence [1,2,3,4,5,6]. The NCI-60 data have been widely used in cancer research and bioinformatics, but the multiple datasets may be most informative for the recognition of complex biosignatures. Such biosignatures may in turn lead to increased understanding of cell phenotypes and pathway relationships within the cell. We previously developed GoMiner [7] and High-Throughput GoMiner [8], applications that organize lists of interesting genes (for example, under- and over-expressed genes from a microarray experiment) for biological interpretation in the context of the Gene Ontology [9,10]. GoMiner and related tools typically generate a list of significant functional categories. In addition to lists and tables, High-Throughput GoMiner can provide two kinds of clustered image maps (CIMs) as graphical output. Integrative categories versus experiments CIMs capture the relationships between categories and multiple experiments; individual categories versus genes CIMs capture the relationships between categories and genes. Both types of CIMs are used to present the results in the present work. In the past decade, systems biology has become increasingly prominent as the numbers of analyzable genes and biological parameters have increased, and is beginning to show their functional relationships. A standard approach for studying systems biology with genomic data is to cluster genes whose expression profiles co-vary either over a time course or across multiple samples. For example, Garraway et al. [11] performed an integrated supervised analysis of SNP array and gene expression data to identify MITF as a lineage survival oncogene amplified in malignant melanoma. A number of additional gene expression microarray demonstrate the potential of gene co-expression studies. For example, Prieto et al. [12] used the Affymetrix HGU133A platform to identify co-expression networks in a diversity of human tissue samples. Their network revealed a map of coexpression clusters organized in well-defined functional constellations. Two major regions in this network corresponded to genes involved in nuclear and mitochondrial metabolism. That study is not directly relevant to cancer, though, since no cancer tissues were included in the study. Choi et al. [13] did study cancer tissues, but had unfortunately culled published data from what would now be considered to be outdated (Affymetrix U95A) or unreliable (cDNA) platforms. Also, the data obtained on different platforms needed to be reconciled, and the date of the studies preceded the availability of reliable resources like AffyProbeMiner [14] and SpliceCenter [15]. Nevertheless, Choi was able to detect functional differences between normal growth and cancer in terms of gene coexpression changes in broad areas of physiology: energy metabolism, the cell cycle, immune activation and collagen production. Other studies have been focused on tissue-specific genes. Cho et al. [16] revealed many pathways related to the pathophysiology of lung cancer: Cytokine Network and TNF/Stress Related Signaling pathway pair; thrombin signaling and protease-activated receptors pathway; Cell Cycle: G1/S Check Point and Inhibition of Cellular Proliferation by Gleevec. Likewise, the studies of Lai et al. [17] were restricted to prostate cancer and developed a statistical method for identifying differential genegene coexpression patterns in different cell states. For a gene of interest, other genes are selected that have differential genegene coexpression patterns with this gene in different cell states. By us (...truncated)