Democratising Knowledge Graphs
Hot Topics
BioCypher is the simplest way to create an AI-enabled knowledge graph for biomedical (or other) tasks. See below and the BioChatter website for more information.
We have also recently published a perspective on connecting knowledge and machine learning to enable causal reasoning in biomedicine, with a particular focus on the currently emerging “foundation models.” You can read it here on arXiv.
Building a knowledge graph for biomedical tasks usually takes months or years. What if you could do it in weeks or days? We created BioCypher to make the process of creating a biomedical knowledge graph easier than ever, but still flexible and transparent. BioCypher is built around the concept of a “threefold modularity”: modularity of data sources, modularity of structure-giving ontology, and modularity of output formats. This design allows for a high degree of flexibility and reusability, rationalising efforts by leveraging the biomedical community.
If you’re new to knowledge graphs and want to familiarise with the concepts that drive BioCypher, we recommend to check out the graphical abstract below and read our paper (self-archived version here on Zenodo, online version at this https link)!
Note
BioCypher is an inclusive community-driven project. If you have any questions, specific needs, or want to contribute to the project, please contact us over on our Zulip channel, on GitHub or via email at sebastian.lobentanzer (at) uni-heidelberg.de.
BioCypher uses a collection of reusable “adapters” for the different sources of biomedical knowledge, which can be flexibly recombined to fit various demands, thus reducing redundant maintenance work through quasi-standardisation. Integrating the controlled vocabularies of ontologies into the process helps to harmonise the data from individual resources and yields a consistent semantic basis for downstream analyses. Through unambiguous and simple “low-code” configuration, a reproducible knowledge graph can be created and shared for every specific task.
Mission Statement
We aim to enable access to versatile and powerful knowledge graphs for as many researchers as possible. Making biomedical knowledge “their own” is often a privilege of the companies and groups that can afford individuals or teams working on knowledge representation in a dedicated manner. With BioCypher, we aim to change that. Creating a knowledge graph should be “as simple as possible, but not any simpler.” To achieve this, we have developed a framework that facilitates the creation of knowledge graphs that are informed by the latest developments in the field of biomedical knowledge representation. However, to make this framework truly accessible and comprehensive, we need the input of the biomedical community. We are therefore inviting you to join us in this endeavour!
Vision Statement
The machine learning models we train are only as good as the data they are trained on. However, most developments today still rely on manually engineered and non-reproducible data processing. We envision a future where the creation of knowledge graphs is as easy as running a script, enabling researchers to build reliable knowledge representations with up-to-date information. We believe that making the knowledge representation process more agile and lifting it to the same level of attention as the process of algorithm development will lead to more robust and reliable machine learning models. We are convinced that this will be a crucial step towards the democratization of AI in biomedicine and beyond.
Connect your Knowledge Graph to Large Language Models
To facilitate the use of knowledge graphs in downstream tasks, we have developed a framework to connect knowledge graphs to large language models. This framework is called BioChatter and is used in our web apps. See the links for more information.