LPTK: a linguistic pattern-aware dependency tree kernel approach for the BioCreative VI CHEMPROT task

Area of Science:

• Biomedical Informatics
• Text Mining
• Computational Biology

Background:

• Accurate identification of chemical-protein interactions (CPIs) from biomedical literature is crucial for drug discovery and understanding biological mechanisms.
• Existing text mining methods often struggle with the complexity and nuances of linguistic patterns in scientific texts.

Purpose of the Study:

• To develop and evaluate a novel linguistic pattern learning method for capturing CPIs.
• To introduce an integrated framework for extracting CPIs using enhanced feature representation.

Main Methods:

• Developed the Linguistic Pattern-Aware Dependency Tree Kernel (LPAD-TK) method for learning interaction patterns.
• Integrated LPAD-TK with a smooth partial tree kernel for CPI extraction.
• Utilized a geometric representation for linguistic probability to optimize feature dimensions and interpretability.

Main Results:

• The proposed framework demonstrated effectiveness in identifying CPIs.
• The method achieved superior performance compared to several existing systems across three diverse datasets (CHEMPROT, Chemical-Disease Relation, Protein-Protein Interaction).
• The feature representation approach enhanced classification sufficiency and provided interpretable contexts.

Conclusions:

• The Linguistic Pattern-Aware Dependency Tree Kernel method offers a robust and efficient approach for extracting chemical-protein interactions from biomedical literature.
• This framework advances text mining capabilities in life sciences by providing interpretable and accurate biological interaction extraction.