A fast document classification algorithm for gene symbol disambiguation in the BITOLA literature-based discovery...

Area of Science:

• Biomedical text mining
• Bioinformatics
• Natural Language Processing

Background:

• Gene symbol disambiguation is crucial for accurate biomedical text mining.
• Many gene symbols overlap with general biomedical concepts, posing a significant challenge.
• Existing methods struggle with the ambiguity of gene symbols in large datasets.

Purpose of the Study:

• To develop an efficient and accurate algorithm for gene symbol disambiguation.
• To improve the performance of biomedical text mining systems by resolving symbol ambiguity.
• To specifically address the challenge of gene symbols that are also common biomedical terms.

Main Methods:

• A two-step approach involving citation domain classification followed by gene symbol detection.
• Utilized ontological information from Medical Subject Headings (MeSH) for domain classification.
• Implemented the algorithm within the BITOLA literature-based discovery support system.

Main Results:

• Achieved a high predictive accuracy of 0.91 for gene symbol disambiguation.
• The algorithm demonstrates high efficiency, capable of processing the entire MEDLINE database in hours.
• Successfully integrated into the BITOLA system for practical application.

Conclusions:

• The proposed domain classification method effectively resolves gene symbol ambiguity.
• This approach significantly enhances the accuracy and efficiency of biomedical literature analysis.
• The algorithm offers a valuable tool for advancing literature-based discovery in biology and medicine.