Computational Analysis of Pathophysiological Mechanisms Based on Pathway Modeling

Abstract

The advent of the big data era poses major challenges to the biomedical domain. First, it is necessary to adopt strategies that integrate and link the heterogeneous resources that contain multiscale and multimodal data in order to fill the existing knowledge gaps. Further, there is a need for developing methods designed not only to interrogate the data but also to interpret and decode the complex world of biology.<br /> In this work, we address the two aforementioned challenges in the domain of pathway knowledge. This thesis presents two ecosystems devised to harmonize and consolidate knowledge from disparate pathway databases, ultimately providing a holistic view of the pathway landscape. Leveraging this integrative effort, we designed a benchmarking study that demonstrates significant impact of database selection in functional enrichment methods and prediction modeling. The results of this work advocate for integrative approaches since our unifying schema has been shown to yield more robust and interpretable results than individual databases and to improve the predictability in modeling tasks. Tangential to these pathway-driven approaches, this work also presents two frameworks devised to identify mechanisms and biomarkers in the neurodegenerative and psychiatric field. The first resource, NeuroMMSig, is the largest inventory of candidate mechanisms for Alzheimer’s and Parkinson’s disease. This manually-curated collection of over 200 computable mechanistic networks emerged as a novel knowledge- based paradigm by laying the ground for the first draft of a mechanism-based taxonomy in both conditions. The second resource, PTSDDB, is a database cataloging biomarker information in the context of post-traumatic stress disorder that opens the door for a future systematic meta-analysis of results reported in literature. Finally, we conclude the thesis with a novel approach that bridges the gap between mechanistic knowledge and patient-level data, paving the way for a mechanism-based stratification of dementia patients.<br /> In summary, this thesis presents novel methodologies for the integration of pathway knowledge. In addition, it introduces new resources and strategies in the context of neurodegenerative and psychiatric disorders. These advances have numerous applications in translational research, ranging from drug discovery to patient stratification.