ROBOKOP KG and KGB: Integrated Knowledge Graphs from Federated Sources

Area of Science:

• Biomedical Informatics
• Knowledge Representation
• Data Integration

Background:

• The proliferation of distributed biomedical data sources via Application Programming Interfaces (APIs) has created an implicit knowledge graph (KG).
• Integrating data across multiple APIs presents significant challenges due to incompatible semantic types, identifier schemes, and data formats.
• Existing systems struggle to effectively consolidate and query this heterogeneous biomedical information.

Purpose of the Study:

• To present ROBOKOP KG, a knowledge graph designed to support biomedical question-answering.
• To introduce the ROBOKOP Knowledge Graph Builder (KGB) for constructing and managing the KG.
• To provide an extensible framework for querying and integrating federated biomedical data sources.

Main Methods:

• Development of the ROBOKOP KG using data from distributed biomedical APIs.
• Implementation of the ROBOKOP Knowledge Graph Builder (KGB) for automated KG construction.
• Design of a framework to handle semantic heterogeneity and data integration challenges.

Main Results:

• Successful construction of the ROBOKOP KG, a unified resource for biomedical knowledge.
• The ROBOKOP KGB provides a scalable solution for building and maintaining the KG.
• The framework facilitates querying over integrated, federated data sources.

Conclusions:

• ROBOKOP KG and KGB offer a robust solution for integrating and querying distributed biomedical data.
• The developed framework effectively addresses semantic and data format incompatibilities.
• This work enhances the accessibility and utility of vast biomedical knowledge for research and applications.