Knowledge Graphs, Network Models and Health Data Science Approaches for Toxicology and Pharmacology

Big Data analytics focus on the collection, modelling, and analysis of large-scale data to identify correlations and relationships, gain new insights as well as to make predictions about possible future outcomes or new facts about the world under investigation. In the health sciences, Big Data can have many different facets and applications, ranging from hospital process optimization, over image classification and personalized medicine to drug development and chemicals safety assessment. However, current Big Data studies in the life sciences are often limited to a small range of data sources and data types due to the diversity and complexity of the available data, standards, and interpretations.

Knowledge Graphs are a highly flexible, link-oriented data structure, which, based on the application of a reasoning engine, allow the inference of new facts about the world under investigation. Knowledge Graphs are built upon a graph data model, which is a schema-free, highly flexible, and modifiable data management model. In addition to classical data retrieval and analytical methodologies, graph-based data models are link and path focused as well as allow the application of network metrics to analyse not only individual data points, but with respect to the whole system.

In this thesis I have investigated the use of graph data models and Knowledge Graphs as data management, data integration and knowledge inference engines for the highly diverse data across the life sciences with a focus on their application to the compound safety and development process. In addition, I developed and collected different network analysis methodologies for the analysis of networks created from molecular data as well as networks contained directly or indirectly in a Knowledge Graph data model or in combination with molecular data.