Inference and validation of cancer gene regulatory networks

While cancer is unmistakably a genetic disease, it is becoming increasingly apparent that it shares many of its essential characteristics with those typical for developmental processes. Indeed, cancer cells can be considered a potential cell type arising by acquiring specific mutations that alter its regulatory state. Underlying such a cancerous phenotype is a specific gene regulatory network (GRN). In this light, gaining mechanistic insight into the structure and dynamics of GRNs is essential when trying to understand oncogenesis and cancer progression. Decoding these networks will be key in the rational development of new cancer therapies. To this end, an integrated approach is needed, combining high-throughput technologies such as next generation sequencing with computational methods and advanced experimental validation tools to explore the GRNs underlying cancer. In this thesis, we have used several integrated approaches to identify and decode regulatory networks underlying two different cancer types. On the one hand we have combined transcriptomics, epigenomics and extensive computational methods with experimental validation to determine the regulatory differences that make up the heterogeneity underlying melanoma. Particularly, we identified two distinct cellular states, each with their own regulatory network and pinpointed essential master regulators for these networks. Additionally, using experimental validation techniques we have shown the relevance of these networks to the biology and potential treatment of melanoma. At the same time we have worked on the development of new tools and methods that allow better prediction and validation of GRNs and the components thereof. With the development of iRegulon, we provided a new computational method that proposes potential master regulators of a set of co-expressed genes. By integrating motif and track discovery, this tool brings motif prediction and network engineering to a new level, and allows biologist to gain a better insight into their data. During my PhD I have been closely involved in the testing and validation of this tool and have exploited it to map p53 related networks. For instance, by using iRegulon on a set of co-expressed genes after p53 activation, we managed to expand the network of p53 even further, identifying a range of novel target genes. Additionally, we generated several hypotheses to answer some crucial questions still much debated today in the field of p53. Building on this work, we have also developed a new experimental method called CHEQ-seq that allows researchers to functionally validate predicted enhancers. Using CHEQ-seq on our previously proposed expanded p53 network, we validated a number of regulatory regions as functional enhancers. The tool also allowed us to work towards a model to better understand and predict p53 binding on a global scale and answer some of the questions previously formulated with iRegulon. All together, this work shows that an integrated approach, combining both high-throughput computational and experimental methods can create a more complete view on genomic regulatory control and provide us with a deeper understanding of the molecular pathology of cancer.nrpages: 202status: publishe