kg2

KG2 is a large biomedical knowledge graph developed as part of the Biomedical Translator initiative. It integrates data from numerous life-science resources into a single, queryable graph to support biomedical reasoning, hypothesis generation, and decision support.

Nodes in KG2 represent biological entities such as genes, proteins, diseases, phenotypes, chemicals, and phenotypic concepts, while edges encode relationships such as associated_with, interplays_with, interacts_with, and treatment_of. The graph emphasizes provenance and evidence, tagging edges and nodes with source metadata to enable traceability and confidence assessment. It is designed to enable cross-source reasoning, linking disparate data types across domains such as genomics, pharmacology, and clinical information.

KG2 has variants, including a larger, more comprehensive version and a curated subset intended for clinical use. Data sources are drawn from public and licensed resources and are refreshed periodically to reflect updates in the biomedical literature and databases.

Access to KG2 is provided via downloads and programmatic interfaces, and it is commonly used in conjunction with reasoning tools such as the ARAX framework to power question answering, knowledge discovery, and hypothesis testing.

Development is collaborative, with governance structures that promote interoperability and openness. Documentation, release notes, and data dumps are maintained by the project, and users can cite sources and provenance when using the graph.

See also: Biomedical Translator, ARAX, knowledge graph.