Abstract:

The recent development in quantitative measurements and access to interaction databases facilitate the construction of detailed molecular interaction maps of cellular processes. Such networks serve as a knowledge-base and being machine readable are amenable to computational analysis. Studying biochemical networks as a non-linear dynamical system is challenging due to a large number of components and complex network structures including feedback/feedforward loops.

In this thesis, I proposed an integrative workflow based on multi-objective optimization function to study large-scale biochemical networks by combining techniques from bioinformatics and systems biology. It integrates heterogeneous sources of biological information with network structure and dynamical systems analysis to unravel mechanisms underlying diseases. To understand the functional role of the transcription factor E2F1 in different traits of cancer, such as drug resistance and epithelial-mesenchymal transition (EMT), we constructed a comprehensive molecular interaction map. The network contains 1,015 nodes and 4,180 interactions. Further, the network topological analysis revealed a large number of feedback and feedforward loops. To make such large, complex network suitable for dynamical systems analysis, the proposed workflow combined its topological properties with high-throughput and biomedical data to identify disease phenotype specific modules, which I refer to as “core-regulatory” networks. Such core networks are smaller compared to parent (large) networks and are amenable for analysis with dynamical systems theory. Using the proposed workflow, I identified core-regulatory networks from the E2F1 map underlying EMT in bladder and breast cancer. I carried out dynamical analyses of the core networks using logic-based models. Using in silico stimulus-response and perturbation experiments, molecular signatures and potential targets were detected for each cancer type. The in silico predictions were validated with patient data and through in vitro experiments.

Moreover, I developed a hybrid modeling framework that combines ordinary differential equation models with logic-based models as a strategy to analyze the dynamics of large-scale non-linear biological systems. Using the proposed hybrid modeling strategy, I simulated the known dynamical features of the E2F1-p73/DNp73-miR205 network in drug resistance for different concentrations of the transcription factor E2F1 and receptor molecules. Further, the results of my model analyses suggest that cancer cells might become independent of growth factors when E2F1 is highly expressed.

This thesis is a contribution to interdisciplinary cancer research, providing a methodology for the analysis of large-scale networks in molecular and cell biology.