- main document
Download (13MB) | Lizenz: Print on Demand
Mammalian signal transduction pathways are highly integrated within extended networks, with crosstalk emerging in space and time. This dynamic circuitry is dependent on changing activity states for proteins and organelles. Network structures govern specificity of cellular responses to external stimuli, including proliferation and cell death. Loss of regulation virtually underlies all disease. However, while the contributions of individual components to phenotype are mostly well understood, systematic elucidation for the emergence or loss of crosstalk and impact on phenotype remains a fundamental challenge in classical biology that can be investigated by systems biology. To that end, we established a mathematical modeling platform, at the interface between experimental and theoretical approaches, to integrate prior literature knowledge with high-content, heterogeneous datasets for the non-intuitive prediction of adaptive signaling events.
In the first part of this work, we investigated high-content microscopy datasets of morphological, bio-energetic and functional features of mitochondria in response to pro- apoptotic treatment in MCF-7 breast cancer cells. Data pretreatment techniques were used to unify the heterogeneous datasets. Using fuzzy logic, we established a generalized data-driven modeling formalism to model signaling events solely based on measurements, capable of high simulation accuracy via non-discrete rule sets. Employing neural networks, a generalized fuzzy logic system, i.e. its rules and membership functions, could be parameterized for each potential signaling interaction. An exhaustive search approach identified models with least error, i.e. the most related signaling events, and predicted a hierarchy of apoptotic events, in which upon activation of pro-apoptotic Bax, mitochondrial fragmentation propagates apoptosis, which is consistent with reported literature. Hence, we established a predictive approach for investigation of protein and organelle interactions utilizing cell-to-cell heterogeneity, a critical source of biologically relevant information.
In the second part of this work, we sought to identify network evolution in the topology of MAPK signaling in the A-375 melanoma cell line. To that end, the modeling method was extended to incorporate temporal and topological structure from phosphorylation profiles of key MAPK intermediates treated with different pharmacological inhibitors and acquired over 96 hours. To increase prediction power, a parameter reduction strategy was developed to identify and fix parameters with lowest contribution to model performance. Therefore, training datasets were bootstrapped and signatures of deviation in flexibility and accuracy were calculated. This novel strategy achieved an optimal set of free parameters. Finally, a reduced multi-treatment model encoding the behavior of the full MAPK dataset was systematically trained to a sequentially increasing subset of time points, enabling time-defined identification of discrepancies in reported vs. acquired network topology. To that end, an objective function for fuzzy logic model optimization was implemented, which accounted for time-defined model training. Analysis led to the identification of emerging discrepancies between model and data at specific time points, thus characterizing a potential network rearrangement upstream of MAPK kinase MEK1, consistent with studies reporting increased resistance to apoptosis exhibited by A-375 melanoma cell line. The approach presented here was successfully benchmarked against a recently published fuzzy-logic-based analysis of signal transduction.
|Supervisor:||Eils, Prof. Dr. Roland|
|Place of Publication:||Heidelberg|
|Date of thesis defense:||30 April 2013|
|Date Deposited:||05 Aug 2013 09:19|
|Faculties / Institutes:||The Faculty of Bio Sciences > Institute of Pharmacy and Molecular Biotechnology|
|Subjects:||004 Data processing Computer science
500 Natural sciences and mathematics
570 Life sciences
600 Technology (Applied sciences)
610 Medical sciences Medicine