BDI-Kit: An AI-powered toolkit for biomedical data harmonization

Area of Science:

• Biomedical Informatics
• Data Science
• Computational Biology

Background:

• Biomedical data availability and advanced analytics offer significant potential for scientific discovery and patient care.
• Data heterogeneity across diverse biomedical datasets presents a substantial barrier to integration and analysis.
• Existing data integration methods and open-source tools struggle to manage the complexity of heterogeneous biomedical data.

Purpose of the Study:

• Introduce the Biomedical Data Integration and Harmonization Toolkit (BDI-Kit) as a solution for complex biomedical data harmonization.
• Provide a flexible and extensible toolkit that supports human-AI collaboration for data integration.
• Enable domain experts to perform data harmonization using natural language and computational pipelines.

Main Methods:

• Developed BDI-Kit, an extensible toolkit featuring a Python API for creating computational harmonization pipelines.
• Integrated an AI-assisted chat interface for natural language-based data harmonization by domain experts.
• Designed BDI-Kit for human-AI collaboration, offering a suite of diverse harmonization methods.

Main Results:

• BDI-Kit successfully addresses the challenge of biomedical data heterogeneity.
• The toolkit offers dual interfaces (Python API and AI chat) catering to different user needs.
• Demonstrated BDI-Kit's capabilities through real-world use cases, showcasing its practical applicability.

Conclusions:

• BDI-Kit simplifies the complex process of biomedical data harmonization.
• The toolkit empowers researchers and practitioners by facilitating effective data exploration.
• BDI-Kit accelerates scientific discovery and clinical research through improved data integration.