JeetVora: /* BiomarkerKB Knowledge Graph */

2026-03-24T18:47:28Z

BiomarkerKB Knowledge Graph

← Older revision		Revision as of 18:47, 24 March 2026
Line 4:		Line 4:

	=== Overview ===		=== Overview ===
	The BiomarkerKG integrates BiomarkerKB data into a Neo4j graph database, exposing relationships such as biomarker–disease associations, specimen types, measurement contexts (via LOINC identifiers), and anatomical locations (via UBERON identifiers). By representing this knowledge as a graph, the BiomarkerKG supports discovery tasks such as identifying biomarkers shared across multiple diseases, finding diseases characterized by multiple biomarkers, and detecting biomarkers measurable in different specimen types or anatomical locations.		The BiomarkerKG integrates BiomarkerKB data into a Neo4j graph database, exposing relationships such as biomarker–disease associations, specimen types, measurement contexts (via LOINC identifiers), and anatomical locations (via UBERON identifiers). The graph structure is based on the Biolink data model, which provides structured relationship types (such as ''associated_with'', ''expressed_in'', and ''related_to'') to describe associations between entities. This design enables queries that go beyond simple co-occurrence to capture how biomarkers change in relation to specific disease states, treatments, or exposures. The initial release of the BKG contains over 300,000 nodes and 1.2 million edges, reflecting both curated biomarker associations and cross-links to CFDE resources. By representing this knowledge as a graph, the BiomarkerKG supports discovery tasks such as identifying biomarkers shared across multiple diseases, finding diseases characterized by multiple biomarkers, and detecting biomarkers measurable in different specimen types or anatomical locations.

	=== Data Representation & Formats ===		=== Data Representation & Formats ===
Line 12:		Line 12:
	* '''OWLNETS''' – a format derived from OWL (Web Ontology Language) that represents ontology-based knowledge as edge lists and node metadata, designed for ingestion into graph databases and machine learning pipelines. Available as dataset BMK_000020.		* '''OWLNETS''' – a format derived from OWL (Web Ontology Language) that represents ontology-based knowledge as edge lists and node metadata, designed for ingestion into graph databases and machine learning pipelines. Available as dataset BMK_000020.

	These formats serve as the basis for constructing the graph: N-Triples are first generated from BiomarkerKB JSON records, then converted to OWLNETS edge list and node metadata files using the nt-owlnets-kg-converter, and finally ingested into Neo4j via a CSV-based Extract-Transform-Load (ETL) pipeline.		These formats serve as the basis for constructing the graph: N-Triples are first generated from BiomarkerKB JSON records, then converted to OWLNETS edge list and node metadata files using the nt-owlnets-kg-converter, and finally ingested into Neo4j via a CSV-based Extract-Transform-Load (ETL) pipeline. The serialization CSV files generated during the build process are available for download with instructions for integrating the BKG with the CFDE Data Distillery Knowledge Graph (see ubkg.docs.xconsortia.org). The graph can be queried using Cypher to identify subgraphs within BiomarkerKB and to integrate it with other compatible knowledgebases.

	=== Integration with External Knowledge Graphs ===		=== Integration with External Knowledge Graphs ===

JeetVora: Created page with " == BiomarkerKB Knowledge Graph == The '''BiomarkerKB Knowledge Graph (BiomarkerKG)''' is a structured, machine-readable representation of the biomarker data contained in BiomarkerKB, organized as a graph of interconnected biological entities and relationships. It is designed to enable complex, multi-hop queries across biomarkers, diseases, anatomical sites, variants, and related biological concepts that go beyond what is possible with traditional tabular data access. =..."

2026-03-24T18:35:21Z

Created page with " == BiomarkerKB Knowledge Graph == The '''BiomarkerKB Knowledge Graph (BiomarkerKG)''' is a structured, machine-readable representation of the biomarker data contained in BiomarkerKB, organized as a graph of interconnected biological entities and relationships. It is designed to enable complex, multi-hop queries across biomarkers, diseases, anatomical sites, variants, and related biological concepts that go beyond what is possible with traditional tabular data access. =..."

New page

== BiomarkerKB Knowledge Graph ==
The '''BiomarkerKB Knowledge Graph (BiomarkerKG)''' is a structured, machine-readable representation of the biomarker data contained in BiomarkerKB, organized as a graph of interconnected biological entities and relationships. It is designed to enable complex, multi-hop queries across biomarkers, diseases, anatomical sites, variants, and related biological concepts that go beyond what is possible with traditional tabular data access.

=== Overview ===
The BiomarkerKG integrates BiomarkerKB data into a Neo4j graph database, exposing relationships such as biomarker–disease associations, specimen types, measurement contexts (via LOINC identifiers), and anatomical locations (via UBERON identifiers). By representing this knowledge as a graph, the BiomarkerKG supports discovery tasks such as identifying biomarkers shared across multiple diseases, finding diseases characterized by multiple biomarkers, and detecting biomarkers measurable in different specimen types or anatomical locations.

=== Data Representation & Formats ===
BiomarkerKB data is exported and distributed in two Semantic Web–compatible formats available from the BiomarkerKB data portal:

* '''RDF N-Triples''' – a standard serialization of RDF (Resource Description Framework) triples, enabling interoperability with SPARQL endpoints and linked data ecosystems. Available as dataset BMK_000019.
* '''OWLNETS''' – a format derived from OWL (Web Ontology Language) that represents ontology-based knowledge as edge lists and node metadata, designed for ingestion into graph databases and machine learning pipelines. Available as dataset BMK_000020.

These formats serve as the basis for constructing the graph: N-Triples are first generated from BiomarkerKB JSON records, then converted to OWLNETS edge list and node metadata files using the nt-owlnets-kg-converter, and finally ingested into Neo4j via a CSV-based Extract-Transform-Load (ETL) pipeline.

=== Integration with External Knowledge Graphs ===
The BiomarkerKG is designed to be integrated into larger biomedical knowledge graph ecosystems:

* '''CFDE Unified Biomedical Knowledge Graph (UBKG):''' The BiomarkerKB data is structured to be compatible with the Data Distillery / UBKG framework, maintained by the NIH Common Fund Data Ecosystem (CFDE). Integration follows the UBKG ETL and build processes, allowing BiomarkerKB nodes and edges to be co-queried with data from other CFDE programs including drug targets, therapeutic agents, and cross-disease treatment connections.
* '''Proto-OKN Knowledge Graph:''' The BiomarkerKG is also integrated into the Proto Open Knowledge Network (Proto-OKN), a national initiative funded by the NSF to build an open, interconnected network of knowledge graphs spanning multiple scientific domains beyond biomedicine, including environmental, social, and geospatial data. Within this broader multi-domain network, BiomarkerKB contributes biomarker knowledge that can be linked to molecular entities such as metabolite-to-protein sequence connections, interacting genes and proteins, and biological pathway and reaction information relevant to disease processes.

=== BiomarkerKG Explorer ===
An interactive web interface for the BiomarkerKG is available at '''[https://biomarker-kg.maayanlab.cloud/ biomarker-kg.maayanlab.cloud]''', developed by the Ma'ayan Laboratory at the Icahn School of Medicine at Mount Sinai. The explorer is built on the open-source Knowledge-Graph-UI framework and provides:

* '''BKG Explorer:''' Visual graph navigation supporting Cypher-based queries to view immediate neighbors of any node and compute shortest paths between two entities.
* '''Downloads:''' Access to downloadable graph dumps and data files.
* '''Tutorials:''' Guided instructions for querying and interpreting the knowledge graph.

The interface is implemented in TypeScript/Next.js and backed by a Neo4j database. Source code is publicly available on GitHub.

=== Links ===
{| class="wikitable"
!Resource
!URL
|-
|BiomarkerKG Explorer
|<nowiki>https://biomarker-kg.maayanlab.cloud/</nowiki>
|-
|GitHub (KG build)
|<nowiki>https://github.com/clinical-biomarkers/Knowledge-Graph</nowiki>
|-
|GitHub (KG UI)
|<nowiki>https://github.com/MaayanLab/Biomarker-KG-UI</nowiki>
|-
|RDF N-Triples dataset (BMK_000019)
|<nowiki>https://data.biomarkerkb.org/BMK_000019</nowiki>
|-
|OWLNETS dataset (BMK_000020)
|<nowiki>https://data.biomarkerkb.org/BMK_000020</nowiki>
|}

BiomarkerKG - Revision history

JeetVora: /* BiomarkerKB Knowledge Graph */