Introduction
Knowledge graphs are gaining popularity for their ability to visually and intuitively represent complex data. People tend to think of entities and connections rather than static plots or tables. Knowledge graphs have proven valuable across different industries, such as retail and finance. Moreover, knowledge graphs have also demonstrated their value in healthcare by aiding drug discovery or handling patient data. This blog post will show how knowledge graphs can help you gain new insights for your biomedical research.
In a recent blog post, The Hyve launched its customized Open Targets Platform (OTP) to showcase the flexibility and power of Open Targets by integrating cBioPortal data. The Hyve’s addition of knowledge graphs to their customized OTP highlights the platform's flexibility while demonstrating how knowledge graphs enhance data exploration.
What is a Knowledge Graph, and Why Should You Use It?
A knowledge graph represents data as entities, their attributes, and the relationships between them. Unlike traditional databases, which store data in tables, knowledge graphs store information as triples (subject-predicate-object). This triple-based storage allows fast execution of complex queries—like path traversals—which enables the retrieval of diverse information quickly and efficiently. This would not be possible when using a table-based (relational) database where you would end up with a very slow query containing multiple ‘join’ clauses. Therefore, Knowledge graphs can reveal patterns or correlations in your data that were not visible before, boosting your research.
Addition of Knowledge Graphs to The Hyve's Open Targets Platform using BioCypher
Like Open Targets, BioCypher is open-source and community-driven, making their integration a natural fit. This shared approach encourages continuous innovation, paving the way for advanced tools to transform our work with biomedical data.
To showcase the strength, The Hyve’s Open Targets experts have implemented a knowledge graph functionality within our customized OTP, using the biomedical framework BioCypher. This framework utilizes the creation of knowledge graphs with the FAIR (Findable, Accessible, Interoperable, Reusable) and TRUST (Transparency, Responsibility, User Focus, Sustainability, Technology) principles in mind. Using a standard framework allows easy expansion with other data sources.
An exciting ongoing development within the BioCypher community is BioChatter, a framework that integrates large language models (LLMs) with knowledge graphs. BioChatter allows users to interact with knowledge graphs using natural language queries, significantly expanding the range of questions that could be asked. This functionality is not yet available in The Hyve’s customized OTP, meaning users can only interact with the knowledge graph using predefined queries. However, all predefined queries can be tailored to the user's needs, and the graph can still be used to the fullest extent.
The Hyve’s Open Targets experts are eager to explore BioChatter's potential in future developments and look forward to collaborating with the community—or perhaps with you!—to bring this cutting-edge functionality to the next level.
Use Case and Biological Insights
Given the significant challenges in treating glioblastoma—such as its aggressive nature, infiltration into surrounding tissue, and limited treatment efficacy—exploring new therapeutic options, including drug repurposing, is crucial. You can visit OTP’s disease profile page to investigate the genes and drugs linked to glioblastoma. Figure 1 illustrates the relationship between targets and drug therapies for Glioblastoma Multiforme (GBM). Notably, several drugs associated with PIK3CA have not yet been linked to glioblastoma, suggesting they could potentially be repurposed for treating this challenging disease. For example, dactolisib, a dual PI3K/mTOR inhibitor, is under investigation for its potential effectiveness in targeting the underlying pathways involved in GBM (Glaviano et al., 2023).
Moreover, investigating biological processes is essential when developing or repurposing a drug for a disease. The OTP contains Gene Ontology (GO) terms annotating biological processes, together with their cellular components and molecular functions. However, these terms are only linked to targets, not to diseases. By incorporating a knowledge graph, we can bridge this gap by identifying the GO terms associated with targets associated with the specific disease of interest. In other words, we can now determine which GO terms are associated with the disease.
Returning to our use case with GBM, Figures 2 and 3 illustrate some key GO terms associated with the disease. The glutathione metabolic process is particularly relevant, as it involves the synthesizing and degradation of glutathione, an important antioxidant that protects cells from oxidative stress—critical in GBM due to its high oxidative environment.
A recent publication highlighted that GBM shows the highest levels of gene expression related to the cysteine and methionine pathways among 32 human cancers by leveraging The Cancer Genome Atlas (Noch et al., 2024). This study also found that N-acetylcysteine (NAC) can induce GBM cell death by reducing mitochondrial oxygen consumption and membrane potential, ultimately leading to mitochondrial cristae dissolution. These effects are amplified under glucose starvation, revealing a unique susceptibility of glioblastoma to cysteine that FDA-approved therapies like NAC may target.
Figures 2 and 3 also highlight the importance of cell adhesion and cytokine response. Disruptions in cell adhesion regulatory mechanisms can facilitate tumor invasion, suggesting that agents and therapies that modulate adhesion could be promising candidates for drug repurposing. Meanwhile, the response to cytokine pathways may indicate changes in immune responses and the tumor microenvironment, further underscoring the potential therapeutic avenues. In further research efforts to develop or repurpose a drug for GBM, consider identifying relevant biological processes using our Open Targets Platform knowledge graph.
Let’s Collaborate!
The use case explained above is only a small example of what you can achieve with a knowledge graph. This new feature is also implemented on drug, target and gene-disease evidence profile pages to support other types of research questions. The Hyve can tailor the knowledge graph feature to fit your specific preferences and needs; more data can be added to the knowledge graph, and the interface can be enriched with features such as filter buttons to aid your use case questions. All of these are to streamline and improve your experience with Open Targets. Don’t hesitate to reach out and explore all the possibilities!
The Hyve's Open Targets Platform is used for demo purposes, showcasing the enormous potential for new features in the Open Targets Platform. While some features, like the knowledge graph implementation, are still in a beta state, we are continuously refining them to unlock the full potential of the available data. We would like to continue the development of the knowledge graph feature—like implementing BioChatter—with every new opportunity that will arise.
Open Targets