Unsupervised gene function extraction using semantic vectors.

Area of Science:

• Biomedical Informatics
• Computational Biology
• Natural Language Processing

Background:

• Identifying gene functions from scientific literature is crucial for biological research.
• Manual curation of gene functions is time-consuming and prone to errors.
• Automated methods can enhance the efficiency and accuracy of gene function extraction.

Purpose of the Study:

• To develop an open-information extraction (open-IE) method for identifying gene function terms (Gene Ontology or GO terms) from biomedical documents.
• To evaluate the proposed method in the BioCreative IV GO shared task (subtask b).

Main Methods:

• The proposed method utilizes distributional semantic similarity over Gene Ontology (GO) terms.
• It is an open-IE approach that does not require annotated training data, promoting generalizability.

Main Results:

• The method achieved an F-measure of 0.26 on the test set in the BioCreative IV GO shared task.
• This result ranked third among seven participating systems.

Conclusions:

• The developed open-IE method demonstrates effectiveness in extracting gene functions from biomedical literature.
• Its reliance on semantic similarity and lack of training data requirement makes it a generalizable approach for bioinformatics research.