Systems Medicine: From Modeling Systems Perturbations to Predicting Drug Synergies
MetadataVis full innførsel
Computational approaches to systems biology and systems medicine enable the study of systems properties that emerge from integrating knowledge about complex interactions, such as in the study of cancer cell phenotypes from cellular signaling networks. This doctoral thesis explores approaches to assemble knowledge of molecular interactions into a comprehensive representation of cellular responses to perturbations. Cellular responses to the gut hormones gastrin and cholecystokinin are analyzed from the perspective of receptor-mediated signaling, demonstrating how current state of the art knowledge can be improved by exploiting a data-driven approach to extending signaling networks. In order to enable reasoning over signaling networks, these are coupled with a logical mathematical formalism, followed by simulation-enabled studies of the response of cell fate networks to pairwise signaling perturbations. A manually curated and parameterized model of the AGS gastric adenocarcinoma cell line correctly predicted 20 of 21 drug combination responses as validated by AGS cell growth experiments. The model correctly identified four synergistic drug interactions. Of these four, two drug synergies described already well-known combination responses that are explored in on-going clinical trials. One of the two novel drug synergies was further validated in in vivo experiments. Based the insights gained from manually curating a logical model the thesis presents foundations on how to automatically obtain predictive models for a given drug panel and a given experimental system. A proof-of-concept approach enabling the automated signaling network assembly and cellular calibration through parameterization of logical equations is presented. This approach can form the basis for a computational pipeline to efficiently generate a well functioning predictive companion model to a given cancer model system. The computational approaches presented here depend on extensive validation in large cancer cell line drug synergy screening experiments. In order to enable efficient identification of experimental synergistic effects the user-friendly and open-source tool CImbinator is presented to analyze and visualize information in such datasets (available at http://cimbinator.bioinfo.cnio.es/).