Translator Curated Query Service
Overview
The Translator Curated Query Service (CQS) is a Standards and Reference Implementation (SRI) service within the NCATS Biomedical Data Translator program. It provides Autonomous Reasoning Agent (ARA)-like capabilities through customizable inference rules, generating predicted biomedical relationships with comprehensive provenance tracking and confidence scoring.
Key Features
- Template-Based Inference: Customizable inference rules captured as CQS templates
- Predicted Edge Generation: Creates inferred relationships from existing knowledge
- Provenance Tracking: Links predictions to supporting auxiliary graphs
- Metadata Attachment: Comprehensive provenance and evidence metadata
- Result Scoring: Confidence scoring for generated predictions
- TRAPI-Compatible: Implements Translator Reasoner API standard
- Query Pattern Support: Handles complex multi-hop reasoning queries
Core Capabilities
Inference Templates
- Customizable Rules: Domain experts define inference patterns
- Pattern Matching: Identifies applicable inference rules for queries
- Multi-Hop Reasoning: Chains multiple inference steps
- Evidence Aggregation: Combines support from multiple knowledge sources
Prediction Generation
- Edge Prediction: Infers new biomedical relationships not explicitly stated in data
- Confidence Assessment: Scores predictions based on evidence strength
- Support Graph Construction: Builds auxiliary graphs showing reasoning chains
- Result Ranking: Orders predictions by confidence and evidence quality
- Source Attribution: Tracks data sources supporting each prediction
- Reasoning Path: Documents the inference steps taken
- Evidence Codes: Standardized codes indicating evidence types
- Confidence Scores: Quantitative measures of prediction reliability
Use Cases
Hypothesis Generation
- Discovering novel drug-disease associations
- Identifying potential drug repurposing opportunities
- Finding mechanistic connections between entities
- Exploring indirect relationships in biomedical knowledge
Knowledge Gap Filling
- Inferring missing relationships from existing knowledge
- Predicting unknown mechanisms of action
- Connecting disparate areas of biomedical knowledge
- Generating testable hypotheses for experimental validation
Clinical Translation
- Supporting precision medicine decisions
- Identifying potential therapeutic interventions
- Understanding disease mechanisms
- Predicting treatment outcomes
Technical Architecture
Query Processing
- Query Reception: Receives TRAPI-formatted queries
- Template Matching: Identifies applicable inference templates
- Rule Application: Applies inference rules to generate predictions
- Graph Construction: Builds auxiliary graphs showing reasoning
- Scoring: Computes confidence scores for predictions
- Response Formatting: Returns TRAPI-compliant results
Template System
- Rule Definition: Expert-curated inference patterns
- Pattern Syntax: Structured representation of reasoning rules
- Condition Specification: Criteria for rule applicability
- Action Definition: Operations to perform when rules match
Integration Points
- Knowledge Providers (KPs): Queries underlying data sources
- Other ARAs: Can be chained with other reasoning services
- ARS: Integrated with Autonomous Relay System
- User Interfaces: Accessible through Translator UIs
This resource has the Information Resource identifier: infores:cqs
Access
- Documentation: https://github.com/NCATSTranslator/Translator-All/wiki/Translator-Curated-Query-Service
- API: Available through Translator infrastructure
- TRAPI Endpoint: Implements standard Translator Reasoner API
For more information about the NCATS Biomedical Data Translator, visit https://ncats.nih.gov/translator