Network approaches to identify cancer drivers from high-dimensional tumor data

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): Large-scale efforts to characterize tumor genomes have uncovered a highly heterogeneous landscape of molecular alterations that distinguish tumor cells from normal cells. A small number of these alterations are causal 'driver' events that confer neoplastic properties to tumors, such as inappropriate growth and proliferation; however, the majority of these alterations are thought to be 'passenger' events that accumulate in tumor cells by chance over the course of tumor progression. Discriminating drivers from passengers is a pressing need in cancer research and will be critical for understanding the molecular origins of tumors, identifying novel targets for drug development, uncovering mechanisms of resistance to therapeutics, and ultimately selecting the most effective therapies for patients. Current efforts t discriminate drivers from passengers rely on statistical over-representation of events in a population of tumors or their predicted effects on protein activity. However, it is now well appreciated that cancer is not a disease of single mutations, nor of genes, but of groups of genes working together in molecular networks and pathways. Cellular behaviors result from complex networks of interactions among biological molecules within the cell, such that driver mutations confer neoplastic behaviors to tumor cells by altering network structure and function. In this grant I propose to model molecular alterations detected in tumors as network perturbations and use these models to discriminate drivers from passengers. These network models will allow us to study cancer in new ways: they will be used to 1) study the biological network effects of known driver mutations versus other human genetic variation, 2) develop hypotheses about the mechanisms by which driver mutations confer neoplastic behaviors to tumor cells, 3) compare patterns of altered network structure across tumor populations, 4) evaluate the combined effect of mutations collocated within a biological network, and 5) predict the set of driver mutations and perturbed pathways in selected individual tumor genomes. I will extend the models to include molecular events overlapping functional non-protein coding elements now known to cover 80% of the human genome, and quantify how acquired alterations observed in a tumor interact with inherited variants in the patient's genome. Finally, will work with established collaborators to experimentally validate novel computational findings uncovered by network perturbation modeling. This project will provide a more global view of the driver landscape in tumors and supply the cancer research community with a suite of computational tools for modeling the consequences of molecular aberrations and targeted interventions in cancer.