A knowledge base for generating patient-specific pathways for individualized treatment decisions in clinical applications
by Florian Johann Auer
Date of Examination:2022-05-18
Date of issue:2022-09-29
Advisor:Prof. Dr. Tim Beißbarth
Referee:Prof. Dr. Tim Beißbarth
Referee:Prof. Dr. Frank Kramer
Persistent Address:
http://dx.doi.org/10.53846/goediss-9457
Files in this item
Name:Dissertation - Florian J. Auer - Revised.pdf
Size:12.8Mb
Format:PDF
The following license files are associated with this item:
Abstract
English
The vast amount of published data on molecular interactions makes it increasingly challenging for researchers and practitioners to find and extract the most relevant information to target the underlying diseases. Network models form a simple and flexible way of representing these complex aspects and associations within biological systems, and their applications are well-established in a wide range of fields in systems medicine. However, biological network data require sophisticated tools for storage, communication, and visualization of relevant information while ensuring proper documentation of performed analyses. Patient-specific results additionally warrant interoperability to existing standards for the application and integration within a clinical setting. Within common bioinformatics workflows data integration, network analysis, and visualization accompany each other and comprise fundamental challenges of combining various tools. The foundation forms the Network Data Exchange (NDEx) and serves as an online commons to store and collaborate on networks of various types. To perform network analyses within the R statistical environment the ndexr package interacts with the NDEx platform and provides the network information using the RCX data model, an adaption to standard R data types and formats compatible with well-established tools for analysis and visualization. Graph Convolutional Neural Networks (Graph-CNN) use retrieved networks in combination with gene expression data as input to predict the metastatic outcome of patients. With the Graph Layerwise Relevance Propagation (GLRP) algorithm the predictions can be attributed to sets of interacting genes thus comprising the patient-specific networks. Visualization of these results significantly contributes to their interpretation and web-based tools thereby facilitate accessibility: NDExEdit allows the application of data-dependent visualizations by mapping network attributes to visual properties and MetaRelSubNetVis focuses on the unambiguous communication of interactive network visualizations in collaborative workflows. Furthermore, an adaption of the patient omics data to the Fast Healthcare Interoperability Resources (FHIR®) standard enables the integration of the results into healthcare systems. Through the combination of biomedical data and their representation through biological networks, it was possible to advance the scientific fields of data integration and visualization where they act as the base for and result of the generation of patient-specific subnetworks at the same time. This contribution to the reproducibility of network-based analyses and interoperability of biological data with clinical systems promotes the integration and application of biological networks within the medical field significantly.
Keywords: data integration; biological networks; network visualization; reproducibilty; interoperability; Fast Healthcare Interoperability Resources (FHIR®); network analysis