Colonet

Colorectal cancer is the second frequent malignancy in Germany and a major cause of cancer death. During the last years combination therapies with non-specific agents have improved overall response rates and survival times. The clinical application of targeted therapies directed against EGFR and VEGFR (Erbitux/Vectibix, Avastin), has significantly improved therapy success. However, predictive markers useful in daily routine are rare and cannot be applied in general. KRAS and BRAF mutations preclude a successful anti-EGFR therapy but response rates hardly approach 30 % even in patients with a seemingly full compliance of molecular requirements, i. e. the wild-type status of the proto-oncogenes. Among many parameters critically influencing therapeutic efficacy are the combinatorial action of receptor tyrosine kinases (e. g. EGFR with GPCR), the impact of underlying genetic and epigenetic alterations determining target gene expression, signaling network composition as well as sex, age, immune status and circadian determinants. This indicates that future diagnostics need to integrate multiple critical parameters to rationally design patient-tailored therapies.

Therefore, the major objective of the ColoNET project is to accomplish an in silico network of genetic, epigenetic and signaling processes to tailor diagnostics with predictive information for therapies targeting receptor tyrosine kinases (e. g. EGFR) in colorectal cancer.

The first step will be to generate a core model consisting of a static, colon-specific signaling network model that serves as a framework for integrating detailed kinetic models of the MAPK, the PI3K and the Wnt pathways. We will include existing genetic and epigenetic information, high-throughput expression data and knowledge on predictive markers obtained by literature mining as well as physiologically relevant data. The dynamics of MAPK, PI3K and Wnt pathways, which are crucial for targeted therapies based on antibodies against tyrosine kinase receptors or small molecule inhibitors, will be subjected to detailed mathematical modeling based on results from focused experimental measurements. Once detailed models of pathways, key signaling nodes, fee-back loops and marker combinations are established, we will turn the static network into a dynamic model in an iterative process. The network model is expected to provide diagnostic guidance for the improvement of therapy response prediction by achieving an in-depth functional understanding of receptor-mediated signaling and downstream processes. We also expect that our model will also be useful for other types of cancer and related predictive diagnostics.