BiomarkerKB Resource Integration

BiomarkerKB collects data from a wide range of resources. Not all collected data are directly integrated into the core data model; some are included as contextual annotations or cross-references to enrich existing entries.

Resources for Exploration

• CADSR Cancer
• Marker Database
• biomarker.org
• ResMarkerDB
• SalivaDB
• GlycanAge Publications
• Cancer Genome Interpreter (Biomarkers)
• Glycan Biomarkers (code)
• Alliance Genome

For suggestions of additional biomarker data resources, please contact: mazumder_lab@gwu.edu

Data Sources

GWAS

Status: Direct integration into data model

• Genome-wide association studies (GWAS) provide biomarkers in the form of SNPs.
• The GWAS Catalog includes SNPs associated with a wide range of diseases.
  • Preliminary curation has only focused on cancer.
  • As of 12/11/2026, biomarkers for all available conditions in the GWAS Catalog have been integrated.
• License: CC BY-NC 4.0

MetaKB

Status: Direct integration into data model

• Provides harmonized associations between cancer genomic variants, diseases, and therapeutic evidence.
• Aggregates and standardizes variant interpretation data from six major knowledgebases:
  • Clinical Interpretation of Variants in Cancer (CIViC) (integrated)
  • OncoKB (restricted from commercial use)
  • The Jackson Laboratory Clinical Knowledgebase (JAX-CKB) (restricted from commercial use and has share-alike requirements for non-commercial use)
  • MolecularMatch (restricted from commercial use)
  • Precision Medicine Knowledgebase) (pending integration)
  • Cancer Genome Interpreter (CGI) – through its Cancer Biomarkers Database component (integrated)
• Enables mapping of:
  • Variant → Disease → Drug relationships
  • Evidence levels and citations
  • Ontology-aligned entities (genes, variants, diseases, drugs)
• Notes:
  • Requires validation of entity mappings against BiomarkerKB schema
• Focused on somatic variant–based biomarkers; contextual attributes such as tissue type, therapy response, or evidence type can be inferred or imputed where not directly specified.
• Manual curation may be required for entries with incomplete evidence annotation or lacking standard ontology references.
• Integration approach: direct mapping of variant, condition, and evidence entities; cross-references retained to original data sources.
• License: Aggregated data are available for non-commercial, research use only, respecting constituent licenses:
  • CIViC – CC0 (Public Domain)
  • PMKB – CC-BY 4.0
  • CGI – CC0 for biomarkers database, CC-BY-NC 4.0 for tool
  • JAX-CKB – CC-BY-NC-SA 4.0
  • OncoKB – custom non-commercial license
  • MolecularMatch – restricted commercial use
  • MetaKB codebase – MIT license
• Overall usage requires adherence to non-commercial research terms; commercial use needs separate permissions from individual data providers.

Glycan LLM Biomarkers

Status: Direct integration into data model

• LangChain LLM method used to collect biomarkers from PubMed Central abstracts
• Method identifies glycan entities and changes mentioned in them associated to disease

Note: only biomarkers with assessed_entity_type: protein were integrated, with the goal of expanding to glycan entity types once the Glycan Structure Dictionary is finalized.

Top 50 Biomarkers

Status: Direct integration into data model

• Biomarkers collected during Summer Volunteership
• Volunteers identified top 50 biomarker entities from BiomarkerKB
• Using this information the top 50 biomarker entities were searched in PubMed
• 100 biomarkers were manually curated

EDRN

Status: Sample integration into data model

• Cancer biomarkers.
• Sample of EDRN Biomarkers provided from EDRN LLM method
• Biomarkers are extracted from free text in EDRN publicly available biomarkers

LOINC

Status: Direct integration into data model

• Metabolite data only
• We are currently working with the Metabolomics Workbench group to get the complete data

OncoKB

Status: Cross-Reference

• Provides useful information on drugs and therapy options for different biomarker entities.
• Also provides information based on what condition the entity is related to.
• License: A license is required to use OncoKB for commercial and/or clinical purposes, and to access OncoKB data programmatically for academic purposes.
• Paid license is required
• Cross-reference from biomarkers in BiomarkerKB to the appropriate drug information and therapy information is the best solution.

HPO

Status: Cross-Reference

• HPO provides disease and entity associations.
• Does not provide a change within the entity so we cannot collect biomarker data from here.
• However we can use it as a cross-reference within our cross-referencing section.
• Provides cross-reference to OMIM, SNOMED, and MONDO.

UniProtKB

Status: Direct integration into data model

• Can provide biomarker (change in entity), entity, condition, and sampling data.
• This data is in a text file that has to be reviewed fully and to make sure it will be able to be automatically extracted.
• Contextual information can be imputed if necessary.
• In UniProt there are found_in and entries that are actual biomarkers:
  • found_in will get a cross-reference;
  • actual biomarkers will be directly integrated.
• Manual curation of 56 reviewed entries with mention of "biomarker" in flat text file.
• License: Creative Commons Attribution 4.0 International (CC BY 4.0).

CIViC

Status: Direct integration into data model

• Clinical Interpretation of Variants in Cancer (CIViC).
• Provides cancer biomarkers in form of DNA mutations (dbSNPs).
• Platform provides clinicians treatment options for patients based on unique tumor profile.
• License: Creative Commons Attribution-NonCommercial 4.0 International License.

ClinVar

Status: Direct integration into data model

• Public archive of reports of human variations classified for diseases and drug responses.
• Provides biomarkers for all disease, but we have only curated cancer biomarkers for now.
  • dbSNPs
  • File is really big but will go back and use existing script to map all biomarkers from here into the data model.
• License: Creative Commons Attribution-NonCommercial 4.0 International License.

Note: Only biomarkers from "cancer" and "carcinoma" tags were pulled. Pending integration of biomarkers for all diseases.

MarkerDB

Status: Cross-Reference

• Provides a lot of useful biomarker data and cross-references other resources as well.
• Information includes: panel information, abnormal levels of biomarkers by disease, structural information, etc.
• Annotations that can be cross-referenced include the above.
• By cross-referencing, BiomarkerKB will allow users to find more information for specific biomarkers and move towards the goal of being a comprehensive resource for biomarkers.
• License: Creative Commons Attribution-NonCommercial 4.0 International License.

Metabolomics Workbench

Status: Direct integration into data model

Data provided by Metabolomics Workbench

• Metabolite biomarkers utilized in the uniform newborn screening program.
• Detect treatable disorders that are life threatening or having long-term morbidity, before they become symptomatic.

OncoMX

Status: Direct integration into data model

• Integrated cancer mutation and expression resource for exploring cancer biomarkers
• Manual curation effort by GWU and JPL
• Over 600 single and panel biomarkers
• License: Creative Commons Attribution-NonCommercial 4.0 International License.

OpenTargets

Status: Direct integration into data model

• Collects potential drug targets and therapeutic targets.
• Some effort was required to find the correct biomarker data.
• 1200 biomarkers collected.
  • dbSNPs related to cancer and other disease
• License: Creative Commons Attribution-NonCommercial 4.0 International License.

Note: Only cancer data was integrated.

PubMed Central Biomarker Gene Set

Status: Direct integration into data model

Data provided by Avi Ma'ayan's LINCS group

• This data set was created through manual curation of biomarker gene sets on Pubmed Central using the results of gene sets returned from Rummagene.
• Using the outputted search results within the Rummagene web server, we manually identified publications that associated different conditions and environmental exposures to biomarker gene sets.
• The biomarker gene sets were retrieved through the validation of the gene mentioned within each of the publications.
• The primary use case for this data is to identify biomarker panels/ gene sets associated with conditions.

SenNet

Status: Direction integration into data model

• Cell senescence biomarkers from SenNet group
• Biomarker data was collected and incorporated however biomarker field was incomplete and data integrated was given a score of -2
• Data is still valuable as contextual data and can be revisited to complete biomarker field in future

For infomation about Cross-references and Annotations in BiomarkerKB please visit - Xrefs and annotations