Federated SPARQL query performance evaluation for exploring disease model mouse: combining gene expression,...

Area of Science:

• Bioinformatics
• Computational Biology
• Biomedical Research

Background:

• RIKEN BioResource MetaDatabase aids bioresource exploration.
• Swiss Institute of Bioinformatics maintains Bgee and OMA databases for gene expression and orthology.

Purpose of the Study:

• Evaluate distributed SPARQL query performance for selecting model organisms in medical research.
• Integrate RIKEN BioResource data with multiple RDF datasets (Bgee, OMA, DisGeNET, MGI, UniProt, ontologies).
• Identify disease-related genes, their expression patterns, and relevant bioresources for specific research applications.

Main Methods:

• Combined RIKEN BioResource data with RDF datasets from Bgee, OMA, DisGeNET, MGI, UniProt, and disease ontologies.
• Utilized SPARQL queries to investigate disease-related genes and their anatomical expression.
• Employed Property Paths for transitive searches of Uberon terms.
• Compared federated SPARQL query performance against a centralized query.

Main Results:

• Identified 14 Alzheimer's disease-related genes (e.g., APP, APOE) expressed in the prefrontal cortex.
• Found 55 RIKEN bioresources (genetically modified mice) relevant to Alzheimer's disease research.
• Identified 14 melanoma-related genes (e.g., HRAS, PTEN) and 12 expressing anatomical parts (e.g., skin of limb).

Conclusions:

• Federated SPARQL query performance degraded.
• Optimized federated query performance by refining transferred data via subqueries.
• Enhanced server specifications to improve triple store query evaluation and reduce performance degradation.