Noodle

Noodle enables semantic discovery and exploration of biomedical literature from PubMed, Europe PMC, and OpenAlex through natural-language research questions, presenting ranked publications with gene, genetic variant, and phenotype views alongside navigable semantic-neighbourhood graphs.

Key Features:

• Natural-language retrieval: Accepts natural-language research questions and identifier-based retrieval for biomedical literature search.
• Multi-source integration: Integrates publication metadata and source-linked context from PubMed, Europe PMC, and OpenAlex while preserving those services as authoritative sources.
• Entity views: Provides gene, genetic variant, phenotype, and indexed-concept views of publications.
• Semantic-neighbourhood graphs: Publication pages include navigable semantic-neighbourhood graphs showing publication relationships.
• Provenance and identifiers: Publication pages provide provenance, identifiers, source links, and indexed concepts.
• SPECTER2 graph: The production SPECTER2 graph covers 1,684,299 canonical non-retracted works with validated 8-12-work neighbourhoods.
• Discovery signals: Uses semantic proximity, citation distance, co-mention, and search rank as discovery signals.
• MCP interface: The companion Noodle Biomedical Literature Discovery MCP exposes the public discovery domain through a read-only, source-linked Model Context Protocol interface with bounded citation and semantic graph traversal.

Scientific Applications:

• Biomedical literature discovery: Supports semantic discovery and navigation of biomedical literature for research exploration.
• Gene and variant exploration: Enables exploration of publications containing gene and genetic variant information.
• Phenotype research: Facilitates identification of publications related to specific phenotypes.
• Publication network analysis: Allows traversal of citation and semantic graphs to explore publication neighbourhoods and relationships.

Methodology:

Noodle integrates publication metadata from PubMed, Europe PMC, and OpenAlex; uses SPECTER2 for generating semantic publication graphs with validated 8-12-work neighbourhoods; and employs semantic proximity, citation distance, co-mention, and search rank as discovery signals.