The Data Distillery: A Graph Framework for Semantic Integration and Querying of Biomedical Data.

Area of Science:

• Biomedical Informatics
• Translational Bioinformatics
• Data Science

Background:

• Biomedical data is often siloed across different domains, hindering integrated analysis.
• Translational research requires linking clinical and experimental data for comprehensive insights.
• Existing data integration frameworks may lack the flexibility for diverse biomedical ontologies.

Purpose of the Study:

• To develop a semantic integration framework for querying heterogeneous biomedical data.
• To support translational research by connecting clinical and experimental datasets.
• To create a unified graph model for the NIH Common Fund Data Ecosystem.

Main Methods:

• Utilized a property graph architecture based on UBKG infrastructure.
• Integrated clinical standards (ICD-10, SNOMED, DrugBank) via UMLS.
• Incorporated genomics and basic science data using ontologies (HPO, GENCODE, Ensembl, STRING, ClinVar).
• Implemented ontology-based ingestion, identifier normalization, and graph-native querying.

Main Results:

• Demonstrated the Data Distillery Knowledge Graph (DDKG) utility across eight informatics use cases.
• Successfully linked diverse datasets including clinical records and experimental results.
• Enabled complex queries for regulatory variant analysis, tissue-specific expression, and biomarker discovery.
• Showcased modularity and extensibility for incorporating new data and schemas.

Conclusions:

• The DDKG provides a robust solution for semantic integration and querying of biomedical data.
• It significantly enhances capabilities for translational research and data-driven discovery.
• The framework is accessible via public interface, API, and downloadable builds.