Normalization of gene/protein names in biological literatures using Vector-Space Model

Area of Science:

• Bioinformatics
• Computational Biology
• Natural Language Processing

Background:

• The exponential growth of biological literature necessitates advanced text mining systems.
• Gene and protein name normalization is crucial for integrating information and building databases or ontologies.
• Existing normalization methods struggle with highly variable gene/protein names found in literature.

Purpose of the Study:

• To propose a novel normalization method for gene and protein names in biological text mining.
• To address the limitations of direct comparison methods in handling name variations.
• To improve the accuracy and efficiency of mapping gene/protein names to databases.

Main Methods:

• Utilizing the Vector-Space Model for gene and protein name normalization.
• Ranking potential database identifiers based on similarity to extracted names.
• Implementing a method to identify the most similar identifier for each gene/protein name.

Main Results:

• The proposed Vector-Space Model method achieved a 70.7% f-measure in experimental results.
• Demonstrated improved performance compared to previous direct comparison methods.
• Successfully handled variational gene/protein names in text mining.

Conclusions:

• The Vector-Space Model offers a robust solution for gene and protein name normalization.
• This method enhances the ability to create comprehensive databases and ontologies from diverse literature.
• The findings contribute to more effective information extraction in bioinformatics.