Matching Biomedical Ontologies through Adaptive Multi-Modal Multi-Objective Evolutionary Algorithm.

Area of Science:

• Biomedical Informatics
• Artificial Intelligence
• Computational Biology

Background:

• Integrating heterogeneous biomedical data is crucial for advancing clinical diagnosis.
• Existing methods for biomedical ontology matching face challenges with large-scale, complex datasets.
• Semantic alignment of biomedical ontologies is essential for data integration.

Purpose of the Study:

• To develop an adaptive Multi-modal Multi-Objective Evolutionary Algorithm (aMMOEA) for matching heterogeneous biomedical ontologies.
• To improve the semantic identification of identical concepts across different biomedical data sources.
• To enhance options for clinical diagnosis through better data integration.

Main Methods:

• Proposed two novel evaluation metrics: f-measure (statistical) and conservativity (logical) for alignment quality.
• Formulated the biomedical ontology matching as a large-scale Multi-modal Multi-objective Optimization Problem (MMOP).
• Developed a problem-specific aMMOEA utilizing a Guiding Matrix (GM) for adaptive convergence and diversity control.

Main Results:

• The aMMOEA effectively addresses the MMOP with sparse Pareto optimal solutions.
• Experimental results on OAEI biomedical tracks demonstrate superior performance compared to state-of-the-art methods.
• The algorithm successfully identifies diverse, semantically equivalent concepts between ontologies.

Conclusions:

• The proposed aMMOEA provides an effective solution for heterogeneous biomedical ontology matching.
• This advancement facilitates better integration of biomedical data, supporting clinical decision-making.
• The adaptive strategy enhances the algorithm's ability to find diverse and relevant semantic alignments.