Gene Name Disambiguation Using Multi-Scope Species Detection

Area of Science:

• Bioinformatics
• Computational Biology
• Text Mining

Background:

• Species detection is crucial in text mining, particularly for gene name disambiguation.
• Existing research often focuses on specific aspects, neglecting a general approach to species detection.
• Identifying the focus species is a key challenge in analyzing biological literature.

Purpose of the Study:

• To develop a general, multi-scope species detection model for biological text mining.
• To improve gene name disambiguation by incorporating species focus detection at different levels.
• To address limitations of previous methods in handling diverse article scopes and implicit species mentions.

Main Methods:

• Developed a multi-scope species detection model identifying focus species at gene mention, sentence, paragraph, and global scopes.
• Utilized species cue words from articles to estimate article-species relevance.
• Proposed the entities frequency-augmented invert species frequency (EF-AISF) formula to calculate relevance scores.
• Introduced a relation guide factor (RGF) for normalizing relevance scores.

Main Results:

• The multi-scope species detection model significantly improved gene name disambiguation.
• The proposed EF-AISF formula effectively estimates the importance of entities to a species.
• The method demonstrated superior performance compared to previous approaches.
• Achieved 88.22% accuracy on the DECA corpus, outperforming prior studies.
• Successfully handled articles without explicit species mentions.

Conclusions:

• The developed multi-scope species detection model offers a generalized solution for species identification in text mining.
• This approach enhances gene name disambiguation and provides a more robust analysis of biological literature.
• The method is effective even for articles lacking explicit species declarations, broadening its applicability.