Constructing Mathematical Models that Predict Therapy Responses in Colorectal Cancer

Biomarkers are the fundamentals of clinical and personalised medicine. Individual diagnosis and treatment choice are dependent on biomarkers. Modern omics technologies, such as genome sequencing, allow molecular profiling of individual patients with extraordinary resolution. However, despite being critical tools in clinical diagnostics, most biomarkers are not mechanistically connected to the disease and based on single timepoint measurements. They, therefore, lack the mechanistic and dynamic information that is needed to follow diseases and therapy. Thus, acquiring an understanding of the dynamic mechanisms of signal transduction networks using critical features from patient-specific mathematical models can help to identify novel potential drug targets. This thesis presents two studies which focus on the WNT signal transduction network for colorectal cancer through a machine learning and mechanistic dynamic approach. Both studies determine prognostic genes and proteins and correctly predict patient-specific differences in signal transduction network activity. The first study, based on a machine learning model, developed three linear-regression models. Each model incorporated both genes and proteins from established WNT mechanistic models in the literature and novel genes with significant associations to event-free patient survival. The second study developed the framework for a mechanistic dynamic model of the entire sequence of events in the WNT signal transduction network, from ligand binding to β-catenin accumulation, and the effects of inhibitors, such as sFRPs and DKK, to the production of WNT target genes. This model, when developed, will have the potential to be tested against tumour profiling and clinical data from patients. The final study of this thesis investigated the ever-growing area of concern; the racial disparities between Black/African Americans and White patients in colorectal cancer. Black/African Americans have a higher incidence of colorectal cancer and worse survival rates when compared with Whites. To this day, there is no specific cause other than possible sociodemographic, socioeconomic, education, nutrition, delivery of healthcare, screening, and cultural factors. Therefore, this study investigated, using a bioinformatic approach, whether genetic factors such as significant genes and critical colorectal cancer signal transduction networks are factors that are contributing to the racial disparity issue. Consequently, this information can guide precision medicine approaches tailored specifically for colorectal cancer racial disparities. Overall, this thesis proposes state-of-the-art computational models of the colorectal cancer WNT signalling network as a new conceptual approach to biomarker discovery and design. By including dynamic information, computational models have the potential to simulate disease evolution and provide therapy responses with high sensitivity and specificity.