Related Experiment Videos
Extraction of protein interaction information from unstructured text using a context-free grammar
Joshua M Temkin1, Mark R Gilder
1GE Global Research, 1 Research Circle, Niskayuna, NY 12309, USA.
Bioinformatics (Oxford, England)
|November 5, 2003
Summary
This study presents a system for automatically extracting protein, gene, and small molecule (PGSM) interactions from scientific literature. Using a context-free grammar, the system achieves high accuracy in identifying PGSMs and their interactions, streamlining data curation for biological databases.
Area of Science:
- Bioinformatics
- Computational Biology
- Natural Language Processing
Background:
- Vast amounts of data on protein, gene, and small molecule (PGSM) interactions are generated in biological research.
- Manual extraction of these interactions from literature is time-consuming and limits database growth.
- Automating interaction extraction is crucial for computational analysis and managing scientific literature.
Purpose of the Study:
- To develop an automated system for extracting PGSM interactions from unstructured text.
- To improve the efficiency and accuracy of populating biological databases with interaction data.
Main Methods:
- Utilized a lexical analyzer and context-free grammar (CFG) for parsing natural language.
- Developed efficient parsers to identify PGSM names and their relationships.
- Focused on domain-specific structure rather than semantic analysis for natural language processing.
Main Results:
- Achieved 83.5% recall and 93.1% precision in recognizing PGSM names.
- Achieved 63.9% recall and 70.2% precision in extracting interactions between PGSMs.
- The CFG approach simplified natural language processing and reduced complexity.
Conclusions:
- The developed system efficiently extracts PGSM interactions from text with high precision and recall.
- This automated approach significantly aids in curating biological databases.
- The system's design allows for adaptation to extract other biological relationships.