NLIMED: Natural Language Interface for Model Entity Discovery in Biosimulation Model Repositories.

Area of Science:

• Computational Biology
• Bioinformatics
• Systems Biology

Background:

• Semantic annotation using Resource Description Framework (RDF) is vital for biosimulation model reusability and reproducibility.
• The Resource Description Framework query language (SPARQL) enables searching model entities but presents challenges due to its complexity and rigid syntax.
• Existing tools like the NCBO Annotator face limitations in effectively identifying relevant ontology classes from natural language queries.

Purpose of the Study:

• To develop NLIMED, an interface that translates natural language queries into SPARQL queries.
• To facilitate efficient querying and discovery of entities within biosimulation models stored in repositories like the Physiome Model Repository (PMR) and BioModels database.
• To improve the accessibility and usability of semantic annotations for biosimulation models.

Main Methods:

• NLIMED employs natural language processing (NLP) techniques to 'chunk' natural language queries into phrases.
• Ontology classes and predicates are identified and annotated using various NLP tools.
• The annotated components are composed into SPARQL queries, ranked using a SPARQL Composer and an indexing system.

Main Results:

• NLIMED demonstrates superior effectiveness compared to the NCBO Annotator in identifying relevant ontology classes from natural language queries.
• The system successfully queries and discovers model entities from the Physiome Model Repository (PMR) and BioModels database.
• NLIMED's approach adapts well to queries related to well-annotated models, as validated against historical query data from the PMR.

Conclusions:

• NLIMED offers a user-friendly interface for querying complex biosimulation model repositories.
• The natural language to SPARQL conversion significantly enhances the discoverability and accessibility of biosimulation model components.
• NLIMED promotes greater reusability and reproducibility of biosimulation models through improved data retrieval.