Automatic, context-specific generation of Gene Ontology slims

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Ontologies like Gene Ontology (GO) are crucial for annotating genome-scale data but have become unwieldy due to their size.
• Current methods for creating GO slims (simplified subsets) are manual, subjective, and time-consuming, requiring expert biological knowledge.

Purpose of the Study:

• To develop an objective, automated framework for generating customized GO slims tailored to specific annotated datasets.
• To improve the efficiency and ease of manipulation of GO for various applications.

Main Methods:

• Developed a framework combining ontology engineering with a data-driven algorithm utilizing graph and information theory.
• Applied the method to Gene Ontology (GO), generating slims at different information thresholds.

Main Results:

• Demonstrated the generation of dataset-specific GO slims using an objective framework.
• Characterized the semantic depth of generated slims and showed increased statistical power in analyses.

Conclusions:

• The developed GO slim creation pipeline is objective, dataset-specific, and available for use with any GO-annotated dataset.
• This automated method is fast, scalable, and applicable to other biomedical ontologies.