Using Semantic Association to Extend and Infer Literature-Oriented Relativity Between Terms.

Area of Science:

• Bioinformatics
• Computational Biology
• Natural Language Processing

Background:

• Relative terms frequently co-occur in scientific literature, prompting methods to weight term relativity and infer new relationships.
• Existing methods for inferring term relationships do not leverage semantic associations within biological ontologies like Gene Ontology and Disease Ontology.

Purpose of the Study:

• To introduce novel methods, Adjusted R-scaled score (ARSS) and ARSS based on information content (ARSSIC), for inferring new term relationships.
• To utilize semantic associations from ontologies to enhance the accuracy of relationship inference in scientific literature.

Main Methods:

• Exploited set inclusion relationships within ontologies to extend term relationships to the literature.
• Developed the ARSS method to measure term relativity across ontologies based on extensional relationships.
• Designed the ARSSIC method, using information shared by term ancestors, to infer new cross-ontology term relationships.

Main Results:

• The ARSS method identified a greater number of statistically significant term pairs compared to other methods, validated by corresponding gene sets.
• The ARSSIC method achieved a high average area under the receiver operating characteristic curve (0.9293), indicating a high true positive rate and low false positive rate.

Conclusions:

• The proposed ARSS and ARSSIC methods effectively infer new relationships between terms by integrating ontology semantic associations.
• ARSSIC demonstrates superior performance in accurately identifying term relationships, offering a valuable tool for literature analysis and knowledge discovery.