NeuroBridge

NeuroBridge applies an ontology-driven deep learning framework and ProvCaRe provenance extraction to generate and rank semantic provenance triples from the biomedical literature to assess and support reproducibility of research findings.

Key Features:

• Ontology-Driven Framework: Uses the S3 model to represent scientific provenance and structure interpretation of study metadata across diverse publications.
• Provenance Metadata Extraction: ProvCaRe employs a provenance-focused text processing workflow to extract semantic triples (subject, predicate, object) from metadata in 435,248 published articles.
• Knowledge Repository: Maintains approximately 48.9 million provenance triples to enable provenance-aware, hypothesis-driven search queries.
• Provenance-Based Ranking Algorithm: Implements a ranking algorithm that prioritizes articles in search results based on the completeness and content of their provenance information.

Scientific Applications:

• Evaluate Reproducibility: Assess the reliability of studies by examining provenance details related to population cohorts, statistical methods, and measurement instruments.
• Enhance Hypothesis Testing: Perform hypothesis-driven searches that emphasize provenance terms to facilitate more robust validation of research hypotheses.
• Identify High-Potential Studies: Detect articles with extensive provenance data across multiple categories that suggest higher likelihoods of reproducibility.

Methodology:

Data modeling using the S3 model to extract semantic provenance, a provenance-focused text processing workflow that generates structured subject–predicate–object triples from article metadata, and repository management maintaining the comprehensive knowledge base and supporting provenance-aware search and provenance-based ranking.