KG based Investigation of Drug Outcome Pairs
Systems pharmacology aims to describe the effects of a drug at the molecular level, by integrating data from multiple temporal and spatial scales across all levels of biological organization. These data are usually represented as a network or knowledge graph (KG), where nodes are biological entities (e.g., chemical compounds, proteins) and edges indicate relationships between these entities (e.g., interactions, drug-target affinity). Systems pharmacology models have successfully predicted drug side effects, developed precision therapeutics, and facilitated drug repurposing. While promising, the majority of these models focus on single entities providing an incomplete “biological milieu”, which may result in biased models and conclusions that are not biologically plausible.
The goal of the proposed work is to demonstrate how a PheKnowLator KG can be traversed and used to construct features capable of discriminating different kinds of drug-outcome pairs.
In order to run the Entity Search you will need to download and run the
To work with the Notebook.
- Fork this library: https://github.com/callahantiff/PheKnowLator
- If the notebook titled
Entity_Search.ipynband accompanying scriptentity_search.pyare not inPheKnowLator/notebooks/tutorials/entity_searchthen download them from the links below toPheKnowLator/notebooks/tutorials/entity_search: - Start the Jupyter Notebook by typing
jupyter notebookin the terminal from thePheKnowLatordirectory
A software demonstration presenting this information will be given at the 2022 Annual Observational Health Data Sciences and Informatics (OHDSI). For more information, including videos and a brief report describing the presented material, please see the OHDSI Symposium page dedicated to this project by clicking on the QR-code provided below.
In 2022, we gave a 10-minute tutorial to the OHDSI Community. For more information, including links to access the video and slides, click here or watch the recorded tutorial below.
