Cancer Systems Biology Centre / Computational Cancer Biology Group NKI

The NWO funded Cancer Systems Biology Center (CSBC) is a centre of excellence within the Netherlands Cancer Institute-Antoni van Leeuwenhoek hospital. The CSBC employs computational modelling and experimental validation cycles spanning multiple levels of complexity including cell lines, mouse models and patients. Within the CSBC several research groups from the NKI-AVL work together to develop a strategy to tackle the complexity of molecular networks that govern breast tumorigenesis with the goal to deliver improved diagnostic tools to enable tailored cancer therapy.

The Computational Cancer Biology group develops novel computational approaches and performs state-of-the-art analyses of a wide array of data types to further our basic understanding of cancer and to translate these findings to the clinic. We follow a three-pronged approach. First, we focus on employing computational approaches to chart the cancer landscape by integrating heterogeneous molecular data sets. Second, we model transcriptional, signalling and metabolic networks to better understand the determinants of the cancer landscape. Finally, we employ patient response data, where available, and drug response data from model systems to map the cancer landscape to (combination) treatments most likely to effectively combat the cancer. Our philosophy is to form strong collaborative ties with experimental and clinical groups such that our computational approaches are firmly embedded in biology and clinical practice.

Theme 1: Charting the molecular landscape of tumors
Recurrent copy number aberration of a single gene is a classical sign of its involvement in tumor development. In many cases tumors become dependent on such driver genes and targeting these genes has proven to be a fruitful strategy in combatting cancer. Similarly, recurrent patterns of copy number aberration involving multiple genes also provide valuable clues regarding ways in which genes collaborate during oncogenesis. For example, genes that are simultaneously inactivated can point to redundant pathways that both need to be inactivated during tumor development. As with single genes, frequent recurrence of such patterns is strong evidence that these patterns were selected for during tumor development and most likely represent dependencies that can be exploited in treatment strategies. Reliably identifying recurrently aberrated (groups of) genes is therefore an essential step in charting the cancer landscape. As a first step, reliable estimates of copy number profiles are required. To this end we have developed an approach to capitalize on the growing amount of (capture-based) DNA sequencing data to accurately estimate copy number profiles of tumors (PropSeg). For the subsequent steps required to chart the landscape, we have developed algorithms to identify recurrently aberrated single genes (ADMIRE) and co-occurrently aberrated gene pairs from copy number data.

Theme 2: Understanding molecular regulatory mechanisms
Understanding the basic molecular regulatory mechanisms that underlie the cancer landscape is important for several reasons. First, it allows more accurate intervention to reach the desired therapeutic effect. Second, it is essential for understanding the functional oncogenic role of spontaneously occurring somatic and germline variants as well as artificially induced perturbations routinely used in genetic screens such as insertional mutagenesis screens. Gene regulatory elements and local chromatin context are indeed important determinants of the effects of mutations on neighboring genes. Similarly, miRNAs are important regulators of target genes, thus impacting the larger genetic network and subsequently oncogenesis and therapy response. We develop Bayesian models of signaling pathways that integrate multiple data types and use measurable variables to estimate unobservable, but informative, variables such as pathway activation. We have applied these approaches on a breast cancer cell line panel exposed to an array of targeted drugs. The models allow the identification of effective (combination) therapies to kill cancer cells.

Theme 3: Mapping the molecular landscape to optimal treatments
In this theme we developed a number of approaches to map molecular profiles of tumors to optimal treatments. Cell line panels provide a rich data source to match molecular profiles of tumor subtypes to drug response. In the most straight-forward approach a tumor subtying is applied to a cell line panel followed by an analysis to determine, for every subtype, the most effective treatment based on drug screening data available for the cell line panel. In collaboration with Astra Zeneca, we have applied this approach on a colorectal cell line panel, and we are also following this approach in collaboration with the Welcome Trust Sanger Institute. In this collaboration, we have access to 1000 cancer cell lines screened across 400 anti-cancer drugs. On this data set we are also developing predictors of therapy response based on logic modeling and integer programming. These predictors map mutation data to drug response and provide easily interpretable models ('Gene A mutated and Gene B mutated predicts Sensitivity') which are highly amenable to the development of testable biological hypotheses and experimental verification. Such models not only provide an effective way to map specific molecular properties (of subtypes) to treatments, but also allow the design of combination treatments as effect modifiers of existing drugs are identified by the models. Within the Cancer Systens Biology Center (CSBC) we develop more focused, detailed models of cellular responses to treatment with the aim of using these models to tailor (combination) treatment. Our efforts are closely linked to the Dutch Center for Personalized Cancer Treatment (www.cpct.nl/en.aspx).