Quality of computationally inferred gene ontology annotations

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Gene Ontology (GO) is the standard for protein function annotation.
• Most GO annotations are electronic and considered unreliable.
• Automated annotations are vital for non-model organisms.

Purpose of the Study:

• To systematically evaluate the quality of electronic GO annotations.
• To assess reliability, specificity, and coverage of automated annotations.
• To develop a method for identifying trustworthy electronic annotations.

Main Methods:

• Utilized changes in successive UniProt GO Annotation database releases.
• Quantitatively assessed electronic annotations based on specificity, reliability, and coverage.
• Developed a methodology for systematic evaluation of automated annotations.

Main Results:

• Electronic GO annotations have significantly improved in recent years.
• Reliability of electronic annotations now rivals curator-inferred annotations (excluding primary literature evidence).
• Over 98% of all GO annotations are inferred without direct curation.

Conclusions:

• A methodology to evaluate and identify reliable electronic GO annotations has been established.
• Electronic annotations are increasingly dependable, especially for non-model organisms.
• This work enables the use of a larger subset of reliable automated annotations.