Matching Biomedical Ontologies via a Hybrid Graph Attention Network

Area of Science:

• Bioinformatics
• Computational Biology
• Ontology Engineering

Background:

• Biomedical ontologies are crucial for organizing complex biological data.
• Independent development leads to heterogeneity and interoperability challenges.
• Existing ontology matching methods rely on labor-intensive feature engineering and miss hidden semantic relationships.

Purpose of the Study:

• To develop an advanced framework for biomedical ontology matching.
• To address the limitations of current feature engineering-based approaches.
• To enhance the interoperability of biomedical knowledge resources.

Main Methods:

• Proposed BioHAN, a hybrid graph attention network framework.
• Implemented an ontology-enriching technique using axioms and external data.
• Utilized hyperbolic graph attention layers for hierarchical concept encoding.
• Employed a graph attention network to aggregate features from neighbors.

Main Results:

• BioHAN effectively refines and enriches ontologies.
• The framework successfully encodes hierarchical concepts in a hyperbolic space.
• BioHAN demonstrates competitive performance against state-of-the-art methods.
• Experimental results confirm the framework's efficacy on real-world biomedical ontologies.

Conclusions:

• BioHAN offers a superior approach to biomedical ontology matching.
• The framework enhances semantic relation capture and interoperability.
• BioHAN represents a significant advancement in biomedical knowledge integration.