arax

is an Aggregator.

It is part of the Translator collection.

ARAX (Expander Agent) is a graph-based modular reasoning tool for translational biomedicine developed as part of the NCATS Biomedical Data Translator program. It provides a web browser user interface and TRAPI-compliant API for encoding translational biomedical questions and integrating knowledge across multiple sources. ARAX features ARAXi, an intuitive domain-specific language for specifying knowledge graph analysis workflows. The system accesses around 40 Knowledge Providers covering over 100 underlying knowledge sources and provides versatile methods for scoring and ranking result subgraphs. ARAX uses RTX-KG2 as its primary knowledge graph and supports query planning, knowledge gathering, overlay of contextual information, filtering, and result ranking through five core modules (Expander, Overlay, Filter, Resultify, and Ranker).

Domains

biomedical, precision medicine, health, translational

License

Warning: No license entered

Homepage

arax

Repository

Unknown

Infores ID

infores:arax

FAIRsharing ID

Unknown

Product Summary

Products

From this Resource
ID Name URL Category Format Description
arax.api ARAX TRAPI API ui ProgrammingInterface ❔ TRAPI-compliant API endpoint for prog...
arax.ui ARAX Web UI arax.ncats.io GraphicalInterface http Web browser interface for querying AR...
arax.araxi.docs ARAXi Documentation DSL_Documentation.md (11.9 KB) DocumentationProduct ❔ ARAXi domain-specific language docume...
arax.github RTX GitHub Repository RTX ProgrammingInterface ❔ Source code repository for ARAX and t...
arax.examples ARAX Examples Examples DocumentationProduct ❔ Example queries and workflows demonst...

Details

Description

ARAX is a computational reasoning tool for translational biomedicine that provides an intuitive system for querying and exploring biomedical knowledge and data. As one of six reasoning engines in the NCATS Biomedical Data Translator system, ARAX enables researchers to formulate complex biomedical questions and integrate knowledge across multiple sources to generate ranked answers.

Key Features

ARAXi Workflow Language

  • Intuitive procedural language for expressing knowledge graph analysis workflows
  • Five core modules: Expander, Overlay, Filter, Resultify, and Ranker
  • Modular design simplifies construction and reuse of workflows
  • Supports incremental query graph construction with add_qnode and add_qedge commands

Knowledge Federation

  • Access to ~40 Knowledge Providers (KPs) covering 100+ underlying knowledge sources
  • Dynamically selects TRAPI-compliant KPs from SmartAPI registry
  • Supports ~135 node categories and 281 edge predicates
  • 76,443 distinct meta-triples (subject-predicate-object combinations)

ARAX Modules

  1. Expander: Queries multiple KPs to build answer knowledge graphs
  2. Overlay: Annotates graphs with contextual quantitative information (Fisher exact test, Jaccard similarity, drug-treats-disease probability, PubMed co-occurrence, clinical data, exposure data)
  3. Filter: Selectively removes nodes/edges based on thresholds
  4. Resultify: Finds subgraphs fulfilling query patterns
  5. Ranker: Scores and ranks results using multi-metric approach

ARAX Node Synonymizer

  • Maps between equivalent concept identifiers from different vocabularies
  • Combines evidence from SRI Node Normalizer, RTX-KG2 same_as relations, identical names, and semantic type compatibility
  • Facilitates seamless integration of results from multiple KPs

Use Cases

Bipolar Disorder Research

Explored connections between D-amino acid oxidase (DAO) and bipolar disorder via protein intermediaries, successfully identifying N-methyl-D-aspartate receptors (NMDARs) in top results.

COVID-19 Mechanism of Action

Identified remdesivir’s mechanism of action through RNA-directed RNA polymerase, revealing “Virus Replication” as a key mediator and potential downstream therapeutic targets like interferon type I.

Drug Repurposing

Link prediction model for drug-treats-disease relationships, successfully distinguishing between existing treats relationships and non-treats edges in RTX-KG2.

Technical Details

Standards Compliance

  • TRAPI (Translator Reasoner API) v1.3+ compliant
  • Biolink Model for semantic layer
  • Uses standardized node categories and edge predicates
  • Compatible with translator ecosystem

Primary Knowledge Graph

  • RTX-KG2 as core KP with 73 knowledge sources
  • Integrates data from ChEMBL, DrugBank, Reactome, OBO Foundry, KEGG, SemMedDB, UniProtKB, UMLS, and more
  • Semantically standardized with Biolink model

Query Input Methods

  1. Interactive visual query graph builder (no coding required)
  2. TRAPI operations (guided entry with drop-downs)
  3. TRAPI JSON (for workflow sharing)
  4. ARAXi workflow language (for complex custom workflows)

Interfaces

Web Browser Interface

  • Interactive query graph builder
  • Visual knowledge graph display
  • Ranked results with detailed provenance
  • Persistent URLs for result retrieval
  • Synonym lookup service
  • System activity monitoring

Programmatic API

  • /query endpoint for synchronous queries
  • /asyncquery endpoint for long-running queries
  • /entity endpoint for node synonymization
  • OpenAPI documentation available

Translator Ecosystem

ARAX is part of the NCATS Biomedical Data Translator, a distributed computational system for accelerating translational science with a layered architecture:

  • Reasoning Engines (ARA): ARAX is one of six reasoning engines
  • Knowledge Providers (KP): ARAX queries ~40 registered KPs
  • Coordinating Services: Interpret and route queries
  • Standard Interfaces: TRAPI for message exchange

Development and Support

Developed by a collaborative team including:

  • Oregon State University
  • Institute for Systems Biology
  • Pennsylvania State University
  • RTX team members

Supported by NCATS through NIH awards OT2TR003428 and OT2TR002520.

Limitations and Future Work

  • System actively developed; no guarantee of indefinite result retention
  • For research purposes only; not for clinical treatment decisions
  • Future enhancements planned:
    • Improved Node Synonymizer accuracy
    • Enhanced drug-treats-disease prediction via alternative embedding models
    • Query backtracking for better pruning strategies
  • RTX-KG2: Primary knowledge graph used by ARAX
  • Biolink Model: Semantic layer specification
  • SmartAPI: Registry of TRAPI-compliant services
  • Other Translator ARAs: ROBOKOP, BioThings Explorer, mediKanren, ARAGORN

Citation

Glen, A. K., Ma, C., Deutsch, E. W., Ramsey, S. A., Koslicki, D., et al. (2023). ARAX: a graph-based modular reasoning tool for translational biomedicine. Bioinformatics, 39(3), btad082. DOI: 10.1093/bioinformatics/btad082

Is this information incorrect or incomplete? Request an update.

Created: October 31, 2025 | Last modified: November 21, 2025