Genome and proteome annotation using automatically recognized concepts and functional networks.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Gene and protein function/disease association prediction is a key area in bioinformatics.
• Current methods primarily use Gene Ontology (GO) terms, overlooking other available ontologies.
• There is a need for broader evaluation of gene-term annotations from diverse ontologies.

Purpose of the Study:

• To broadly evaluate novel, automatically retrieved gene-term annotations from publicly available ontologies.
• To identify concepts within these ontologies that can be predicted using a generalized annotation prediction tool.
• To assess the utility of both manually curated and automatically recognized terms for predictive modeling.

Main Methods:

• Development and application of a generalized tool for predicting gene-term annotations.
• Evaluation of terms from over 250 publicly available ontologies.
• Statistical assessment of term significance using randomly generated gene sets.

Main Results:

• Identified over 127,000 statistically significant terms that can be predicted for human genes.
• Demonstrated that terms from diverse ontologies, not just GO, can be effectively used for prediction.
• Showcased that both manually curated and automatically recognized terms can build robust predictive models.

Conclusions:

• A generalized prediction tool can effectively identify and utilize gene-term annotations across a wide range of ontologies.
• Expanding annotation prediction beyond GO significantly increases the number of predictable terms for human genes.
• This approach enhances the potential for discovering gene functions and disease associations through broader ontological data mining.