Knowledge Graphs in Health: Difference between revisions

From Research Wiki
← Older edit
Fixed doubled title
Added cross-references: Ontologies for Longitudinal Health Records
 
(One intermediate revision by the same user not shown)
Line 1: Line 1:
__NOTOC__
__NOTOC__
The use of graph databases, semantic technologies, and knowledge graphs for health data integration, clinical decision support, and longitudinal patient records.
The use of graph databases, semantic technologies, and knowledge graphs for health data integration, clinical decision support, and longitudinal patient records.


== Overview ==
Neo4j's research shows healthcare leads in knowledge graph adoption, with CIOs unable "to scale AI without a knowledge graph foundation."<ref>Neo4j (2025): "Why Healthcare CIOs Can't Afford to Scale AI Without a Knowledge Graph Foundation" — https://neo4j.com/blog/healthcare-knowledge-graph-ai/</ref> Towards Data Science notes healthcare leads other industries in this space.<ref>Towards Data Science: "Why Healthcare Leads in Knowledge Graphs" — https://towardsdatascience.com</ref>
Knowledge graphs are increasingly recognized as a foundational technology for healthcare AI. Neo4j's research shows "healthcare leads in knowledge graphs" and CIOs "can't afford to scale AI without a knowledge graph foundation."<ref>Neo4j (2025): "Why Healthcare CIOs Can't Afford to Scale AI Without a Knowledge Graph Foundation" — neo4j.com</ref> Towards Data Science notes that healthcare leads other industries in knowledge graph adoption.<ref>Towards Data Science: "Why Healthcare Leads in Knowledge Graphs"</ref>


== Key Projects ==
== Major Projects ==


=== AIDAVA (EU) ===
=== AIDAVA (EU Horizon Europe) ===
[[AIDAVA]] — €7.7M Horizon Europe project (2022–2026) creating a virtual assistant for automated curation and publishing of personal health data. Uses Personal Health Knowledge Graphs (PHKG) based on SNOMED, FHIR, and LOINC ontologies. 13 partners from 9 countries. Coordinator: [[Maastricht University]]. Knowledge graph technology by [[Ontotext]].<ref>aidava.eu / OpenAIRE grant 101057062</ref>
[[AIDAVA]] — €7.7M (2022–2026). Creates a virtual assistant for automated curation and publishing of personal health data using Personal Health Knowledge Graphs (PHKG) based on SNOMED, FHIR, and LOINC ontologies. 13 partners from 9 countries. Coordinator: [[Maastricht University]]. Knowledge graph technology by [[Ontotext]].<ref>European Commission CORDIS, Grant 101057062 — https://cordis.europa.eu/project/id/101057062</ref>


=== AllegroGraph (Franz Inc.) ===
=== OHDSI / OMOP Common Data Model ===
Franz Inc.'s '''AllegroGraph''' is a semantic graph database widely used in life sciences and healthcare. Supports RDF, SPARQL, and knowledge graph applications for clinical data integration and linked health data.<ref>Franz Inc. — franz.com/agraph</ref>
The '''OHDSI''' (Observational Health Data Sciences and Informatics) community uses the '''OMOP Common Data Model''' for observational health research across 100+ data partners globally. Recent work includes graph visualization and validation of drug mappings using LLMs and the USAGI tool.<ref>medRxiv (2025): "Rx Norm for Europe — Toward the representation of medicinal products in the OMOP CDM" — https://www.medrxiv.org</ref> The Jackalope Plus tool enables post-coordination and ontology development for observational studies.<ref>Nature (2025): "Jackalope Plus tool for post-coordination, ontology development" — https://www.nature.com</ref>


=== Neo4j Health Applications ===
=== TriNetX ===
Neo4j, the leading graph database, has published extensively on healthcare knowledge graphs including '''MedQGraph''' (Temporal Medical Knowledge Graphs for AI-Driven Healthcare Insights) and clinical data integration patterns.<ref>Neo4j — neo4j.com, MedQGraph</ref>
'''TriNetX''' operates a global federated real-world data network connecting healthcare organizations. Recently acquired Zetta Genomics to enhance federated genomics data capabilities across its provider network.<ref>BeBeez International (2025): "TriNetX Enhances its Ability to Federate Genomics Data" — https://www.bebeez.it</ref> Also launched NLP services for clinical trial protocol design.<ref>Tech Networks (2025): "TriNetX Unveils Natural Language Processing Service" — https://www.technetworks.com</ref>


=== GraphDB (Ontotext) ===
=== NIH All of Us Research Program ===
'''GraphDB''' by [[Ontotext]] (now part of Graphwise) is an enterprise knowledge graph database used in health data integration. Powers the knowledge graph backbone of AIDAVA.<ref>PR Newswire (2024): "Semantic Web Company and Ontotext Merge to Create Knowledge Graph and AI Powerhouse Graphwise"</ref>
The '''All of Us''' research program (NIH) collects health data from 1M+ diverse US participants. Plans to add environmental exposures data, building toward a comprehensive longitudinal health knowledge graph.<ref>NIH/NIEHS (2025): "'All of Us' study plans to add environmental exposures" — https://www.niehs.nih.gov</ref>


=== OHDSI / OMOP Common Data Model ===
=== UK Biobank Knowledge Graph ===
The '''OHDSI''' (Observational Health Data Sciences and Informatics) community uses the '''OMOP Common Data Model''' with graph-based approaches for observational health research. Recent work includes graph visualization and validation of drug mappings using LLMs.<ref>medRxiv (2025): "Rx Norm for Europe — Toward the representation of medicinal products in the OMOP CDM"</ref>
'''UK Biobank''' — 500,000 participant biomedical database. AI-powered knowledge graphs link cardiac imaging to genomics and drug predictions.<ref>Medical Xpress (2025): "AI-powered knowledge graph links heart images to genes and drug predictions" — https://medicalxpress.com</ref>


=== PubMed Knowledge Graph 2.0 ===
=== PubMed Knowledge Graph 2.0 ===
Connects papers, patents, and clinical trials in biomedical science using graph structures.<ref>Nature (2025): "PubMed knowledge graph 2.0"</ref>
Connects papers, patents, and clinical trials in biomedical science using graph structures. Published in Nature.<ref>Nature (2025): "PubMed knowledge graph 2.0" — https://www.nature.com</ref>
 
=== OpenEHR ===
'''OpenEHR''' is an open standard for clinical health data with archetype-based models. Research shows knowledge graph applications for COVID-19 diagnosis prediction and infection control systems.<ref>Nature (2023): "Prediction of COVID-19 diagnosis based on openEHR artefacts" — https://www.nature.com</ref>
 
=== GA4GH ===
The '''Global Alliance for Genomics and Health''' (GA4GH) develops data standards for genomics. New platforms make genome analyses scalable and reproducible using graph-based approaches.<ref>Berlin Institute of Health (2025): "New platform makes genome analyses scalable and reproducible" — https://bih-charite.de</ref>
 
== Technology Platforms ==
 
=== Graph Databases ===
* '''Neo4j''' — leading graph database, extensive healthcare applications including MedQGraph (Temporal Medical Knowledge Graphs for AI-Driven Healthcare Insights). Snowflake now supports Neo4j graph analytics natively.<ref>Neo4j (2025): "MedQGraph: TMKGs for AI-Driven Healthcare Insights" — https://neo4j.com</ref>
* '''GraphDB (Ontotext)''' — enterprise knowledge graph database, powers AIDAVA. Now part of Graphwise after merger with Semantic Web Company.<ref>PR Newswire (2024): "Semantic Web Company and Ontotext Merge to Create Graphwise" — https://www.prnewswire.com</ref>
* '''AllegroGraph (Franz Inc.)'''' — semantic graph database widely used in life sciences. Supports RDF, SPARQL, and knowledge graph applications for clinical data integration.<ref>Franz Inc.: "AllegroGraph" — https://franz.com/agraph</ref>
* '''Amazon Neptune''' — managed graph database, healthcare analytics use cases.<ref>AWS (2025): "Building health care agents using Amazon Bedrock AgentCore" — https://aws.amazon.com</ref>
* '''Stardog''' — enterprise knowledge graph platform used in life sciences and pharma for data virtualization and reasoning.<ref>Databricks (2025): "Democratize insight with generative AI and knowledge graphs" — https://www.databricks.com</ref>
* '''ArangoDB''' — multi-model database with graph capabilities used in healthcare analytics.<ref>MarketsandMarkets (2025): "Knowledge Graph Market Report 2025-2030" — https://www.marketsandmarkets.com</ref>
* '''TigerGraph''' — scalable graph analytics for clinical data.<ref>Frontiers (2025): "A solution for combining multi-source heterogeneous data to construct enterprise knowledge graph" — https://www.frontiersin.org</ref>


=== i-HD Projects ===
=== Semantic Platforms ===
[[i-HD]] participates in multiple projects using knowledge graph approaches for health data, including IDERHA (integration of heterogeneous data), QUANTUM (data quality labels), and SYNTHEMA (synthetic data generation).<ref>i-hd.eu/projects</ref>
* '''PoolParty (Semantic Web Company / Graphwise)''' — semantic AI platform for health data classification and ontology management. Merged with Ontotext in 2024.<ref>TheRecursive (2024): "Graphwise — A Content Management Platform and a Graph Database Come Together" — https://therecursive.com</ref>
* '''Elsevier Knowledge Graphs''' — life sciences knowledge graphs used for drug repurposing, linking literature, patents, and clinical data.<ref>Elsevier (2025): "How knowledge graphs can supercharge drug repurposing" — https://www.elsevier.com</ref>


=== LLM + Knowledge Graphs ===
=== LLM + Knowledge Graphs ===
Large language models are increasingly used to construct and query health knowledge graphs. Nature published on "Large language model powered knowledge graph construction for mental health exploration."<ref>Nature (2025)</ref> Frontiers published on "Unlocking electronic health records: a hybrid graph RAG approach to safe clinical AI."<ref>Frontiers (2025)</ref>
Large language models are increasingly used to construct and query health knowledge graphs:
* Nature: "Large language model powered knowledge graph construction for mental health exploration"<ref>Nature (2025) — https://www.nature.com</ref>
* Frontiers: "Unlocking electronic health records: a hybrid graph RAG approach to safe clinical AI for patient QA"<ref>Frontiers (2025) — https://www.frontiersin.org</ref>
* Anthropic: "Advancing Claude in healthcare and the life sciences"<ref>Anthropic (2025) — https://www.anthropic.com</ref>
* WHO: "An epidemiological knowledge graph extracted from Disease Outbreak News"<ref>Nature (2025) — https://www.nature.com</ref>
 
=== FHIR + Graph Integration ===
* '''SMART on FHIR''' — open standards platform enabling third-party apps to access clinical data via FHIR APIs. Knowledge graph integration enables semantic queries across patient records.<ref>OCNJ Daily (2026): "10 Best FHIR Software Development Companies" — https://ocnjdaily.com</ref>
* '''HL7 FHIR Bulk Data''' — enabling population-level data export for research. "What Bulk FHIR can do for quality measurement and more."<ref>Healthcare IT News (2025) — https://www.healthcareitnews.com</ref>
* MedCity News: "Beyond the Buzzword: Why Semantic Interoperability is the Holy Grail for Digital Health"<ref>MedCity News (2025) — https://medcitynews.com</ref>


== Technology Stack ==
== Technology Stack ==
Common technologies used in health knowledge graphs:
Common technologies used in health knowledge graphs:
* '''Graph Databases''': Neo4j, GraphDB (Ontotext), AllegroGraph (Franz Inc.), Amazon Neptune
* '''Graph Databases''': Neo4j, GraphDB (Ontotext), AllegroGraph (Franz Inc.), Amazon Neptune, Stardog, TigerGraph, ArangoDB
* '''Standards''': RDF, OWL, SPARQL, SHACL
* '''Standards''': RDF, OWL, SPARQL, SHACL, Property Graph
* '''Health Ontologies''': SNOMED CT, LOINC, HL7 FHIR, ICD-10, OMOP CDM
* '''Health Ontologies''': SNOMED CT, LOINC, HL7 FHIR, ICD-10/11, OMOP CDM, OpenEHR archetypes
* '''Integration''': ETL pipelines, FHIR APIs, NLP extractors
* '''Integration''': ETL pipelines, FHIR APIs, SMART on FHIR, Bulk FHIR, NLP extractors
* '''AI/ML''': Graph neural networks, LLMs for KG construction, embedding models
* '''AI/ML''': Graph neural networks, LLMs for KG construction, embedding models, RAG over graphs


== See Also ==
== See Also ==
* [[Ontologies for Longitudinal Health Records]]
* [[AIDAVA]]
* [[AIDAVA]]
* [[Interoperability]]
* [[Interoperability]]

Latest revision as of 17:57, 14 April 2026

The use of graph databases, semantic technologies, and knowledge graphs for health data integration, clinical decision support, and longitudinal patient records.

Neo4j's research shows healthcare leads in knowledge graph adoption, with CIOs unable "to scale AI without a knowledge graph foundation."[1] Towards Data Science notes healthcare leads other industries in this space.[2]

Major Projects[edit]

AIDAVA (EU Horizon Europe)[edit]

AIDAVA — €7.7M (2022–2026). Creates a virtual assistant for automated curation and publishing of personal health data using Personal Health Knowledge Graphs (PHKG) based on SNOMED, FHIR, and LOINC ontologies. 13 partners from 9 countries. Coordinator: Maastricht University. Knowledge graph technology by Ontotext.[3]

OHDSI / OMOP Common Data Model[edit]

The OHDSI (Observational Health Data Sciences and Informatics) community uses the OMOP Common Data Model for observational health research across 100+ data partners globally. Recent work includes graph visualization and validation of drug mappings using LLMs and the USAGI tool.[4] The Jackalope Plus tool enables post-coordination and ontology development for observational studies.[5]

TriNetX[edit]

TriNetX operates a global federated real-world data network connecting healthcare organizations. Recently acquired Zetta Genomics to enhance federated genomics data capabilities across its provider network.[6] Also launched NLP services for clinical trial protocol design.[7]

NIH All of Us Research Program[edit]

The All of Us research program (NIH) collects health data from 1M+ diverse US participants. Plans to add environmental exposures data, building toward a comprehensive longitudinal health knowledge graph.[8]

UK Biobank Knowledge Graph[edit]

UK Biobank — 500,000 participant biomedical database. AI-powered knowledge graphs link cardiac imaging to genomics and drug predictions.[9]

PubMed Knowledge Graph 2.0[edit]

Connects papers, patents, and clinical trials in biomedical science using graph structures. Published in Nature.[10]

OpenEHR[edit]

OpenEHR is an open standard for clinical health data with archetype-based models. Research shows knowledge graph applications for COVID-19 diagnosis prediction and infection control systems.[11]

GA4GH[edit]

The Global Alliance for Genomics and Health (GA4GH) develops data standards for genomics. New platforms make genome analyses scalable and reproducible using graph-based approaches.[12]

Technology Platforms[edit]

Graph Databases[edit]

  • Neo4j — leading graph database, extensive healthcare applications including MedQGraph (Temporal Medical Knowledge Graphs for AI-Driven Healthcare Insights). Snowflake now supports Neo4j graph analytics natively.[13]
  • GraphDB (Ontotext) — enterprise knowledge graph database, powers AIDAVA. Now part of Graphwise after merger with Semantic Web Company.[14]
  • AllegroGraph (Franz Inc.)' — semantic graph database widely used in life sciences. Supports RDF, SPARQL, and knowledge graph applications for clinical data integration.[15]
  • Amazon Neptune — managed graph database, healthcare analytics use cases.[16]
  • Stardog — enterprise knowledge graph platform used in life sciences and pharma for data virtualization and reasoning.[17]
  • ArangoDB — multi-model database with graph capabilities used in healthcare analytics.[18]
  • TigerGraph — scalable graph analytics for clinical data.[19]

Semantic Platforms[edit]

  • PoolParty (Semantic Web Company / Graphwise) — semantic AI platform for health data classification and ontology management. Merged with Ontotext in 2024.[20]
  • Elsevier Knowledge Graphs — life sciences knowledge graphs used for drug repurposing, linking literature, patents, and clinical data.[21]

LLM + Knowledge Graphs[edit]

Large language models are increasingly used to construct and query health knowledge graphs:

  • Nature: "Large language model powered knowledge graph construction for mental health exploration"[22]
  • Frontiers: "Unlocking electronic health records: a hybrid graph RAG approach to safe clinical AI for patient QA"[23]
  • Anthropic: "Advancing Claude in healthcare and the life sciences"[24]
  • WHO: "An epidemiological knowledge graph extracted from Disease Outbreak News"[25]

FHIR + Graph Integration[edit]

  • SMART on FHIR — open standards platform enabling third-party apps to access clinical data via FHIR APIs. Knowledge graph integration enables semantic queries across patient records.[26]
  • HL7 FHIR Bulk Data — enabling population-level data export for research. "What Bulk FHIR can do for quality measurement and more."[27]
  • MedCity News: "Beyond the Buzzword: Why Semantic Interoperability is the Holy Grail for Digital Health"[28]

Technology Stack[edit]

Common technologies used in health knowledge graphs:

  • Graph Databases: Neo4j, GraphDB (Ontotext), AllegroGraph (Franz Inc.), Amazon Neptune, Stardog, TigerGraph, ArangoDB
  • Standards: RDF, OWL, SPARQL, SHACL, Property Graph
  • Health Ontologies: SNOMED CT, LOINC, HL7 FHIR, ICD-10/11, OMOP CDM, OpenEHR archetypes
  • Integration: ETL pipelines, FHIR APIs, SMART on FHIR, Bulk FHIR, NLP extractors
  • AI/ML: Graph neural networks, LLMs for KG construction, embedding models, RAG over graphs

See Also[edit]

  1. ↑ Neo4j (2025): "Why Healthcare CIOs Can't Afford to Scale AI Without a Knowledge Graph Foundation" — https://neo4j.com/blog/healthcare-knowledge-graph-ai/
  2. ↑ Towards Data Science: "Why Healthcare Leads in Knowledge Graphs" — https://towardsdatascience.com
  3. ↑ European Commission CORDIS, Grant 101057062 — https://cordis.europa.eu/project/id/101057062
  4. ↑ medRxiv (2025): "Rx Norm for Europe — Toward the representation of medicinal products in the OMOP CDM" — https://www.medrxiv.org
  5. ↑ Nature (2025): "Jackalope Plus tool for post-coordination, ontology development" — https://www.nature.com
  6. ↑ BeBeez International (2025): "TriNetX Enhances its Ability to Federate Genomics Data" — https://www.bebeez.it
  7. ↑ Tech Networks (2025): "TriNetX Unveils Natural Language Processing Service" — https://www.technetworks.com
  8. ↑ NIH/NIEHS (2025): "'All of Us' study plans to add environmental exposures" — https://www.niehs.nih.gov
  9. ↑ Medical Xpress (2025): "AI-powered knowledge graph links heart images to genes and drug predictions" — https://medicalxpress.com
  10. ↑ Nature (2025): "PubMed knowledge graph 2.0" — https://www.nature.com
  11. ↑ Nature (2023): "Prediction of COVID-19 diagnosis based on openEHR artefacts" — https://www.nature.com
  12. ↑ Berlin Institute of Health (2025): "New platform makes genome analyses scalable and reproducible" — https://bih-charite.de
  13. ↑ Neo4j (2025): "MedQGraph: TMKGs for AI-Driven Healthcare Insights" — https://neo4j.com
  14. ↑ PR Newswire (2024): "Semantic Web Company and Ontotext Merge to Create Graphwise" — https://www.prnewswire.com
  15. ↑ Franz Inc.: "AllegroGraph" — https://franz.com/agraph
  16. ↑ AWS (2025): "Building health care agents using Amazon Bedrock AgentCore" — https://aws.amazon.com
  17. ↑ Databricks (2025): "Democratize insight with generative AI and knowledge graphs" — https://www.databricks.com
  18. ↑ MarketsandMarkets (2025): "Knowledge Graph Market Report 2025-2030" — https://www.marketsandmarkets.com
  19. ↑ Frontiers (2025): "A solution for combining multi-source heterogeneous data to construct enterprise knowledge graph" — https://www.frontiersin.org
  20. ↑ TheRecursive (2024): "Graphwise — A Content Management Platform and a Graph Database Come Together" — https://therecursive.com
  21. ↑ Elsevier (2025): "How knowledge graphs can supercharge drug repurposing" — https://www.elsevier.com
  22. ↑ Nature (2025) — https://www.nature.com
  23. ↑ Frontiers (2025) — https://www.frontiersin.org
  24. ↑ Anthropic (2025) — https://www.anthropic.com
  25. ↑ Nature (2025) — https://www.nature.com
  26. ↑ OCNJ Daily (2026): "10 Best FHIR Software Development Companies" — https://ocnjdaily.com
  27. ↑ Healthcare IT News (2025) — https://www.healthcareitnews.com
  28. ↑ MedCity News (2025) — https://medcitynews.com
Retrieved from "https://wiki.stamen.health/index.php?title=Knowledge_Graphs_in_Health&oldid=460"