Introduction
Technical background: Graphs to knowledge graphs
Knowledge graph applications
Knowledge graph applications in healthcare
Healthcare Use Case | Graph and Knowledge Graph Mechanics: Explanation and examples |
---|---|
Drug repurposing, Comorbid risk prediction | Link prediction: Prediction of the likelihood of an existing edge between two nodes based on the entirety of the knowledge. For example, prediction of edge likelihood of drug compound and patient to predict personal risk of adverse reaction or links between two diseases posing a high comorbidity risk [7, 17]. |
Disease subtyping | Community detection/graph clustering: Identification of highly connected regions within a real-world data graph that can identify patients with a high similarity, e.g., patients with a certain disease subtype [8]. |
Outcome, status, and risk prediction | Node classification: Prediction of the likelihood of a patient node being assigned a label based on the entirety of their medical data. For example, patient node gets assigned a disease risk label [4]. |
Visual insights | Graph layout and visualization: There have been several studies into the visualization and lay outing of graph-structured data, e.g., biomedical or healthcare data, for aiding human interpretability and pattern recognition [18]. |
Complex patient data queries | Graph traversal: The inherent connected representation in graphs allows for the easy traversal of the graph to identify pieces of information that are separated by several nodes. When combining patient data with terminological knowledge, this allows for complex queries, e.g., identification of all patients based on a medical condition and its subtypes [19]. |