Cross-species gene normalization by species inference.

Area of Science:

• Bioinformatics
• Computational Biology
• Natural Language Processing

Background:

• Accurate gene normalization is crucial for accessing biomedical literature.
• Gene names are often ambiguous due to cross-species references.
• Species information is frequently absent, complicating gene normalization.

Purpose of the Study:

• To improve the accuracy of gene normalization, particularly when species information is not provided.
• To develop an integrated method for gene name recognition, species assignment, and species-specific gene normalization.

Main Methods:

• Propose GenNorm, an integrated method with three modules.
• Address gene name recognition, species assignation, and species-specific normalization.
• Emphasize species assignation as critical for reducing ambiguity.

Main Results:

• GenNorm achieved top-1 threshold average precision (TAP-k) scores of 0.3297 (k=5), 0.3538 (k=10), and 0.3535 (k=20) on difficult test articles.
• In the silver-standard-507 evaluation, GenNorm achieved TAP-k scores of 0.4591 (k=5, 10, 20), ranking 2nd, 2nd, and 3rd respectively.

Conclusions:

• The proposed GenNorm method effectively improves gene normalization accuracy.
• Accurate species assignation is key to resolving gene name ambiguity.
• GenNorm provides a valuable tool for biomedical literature analysis.