Measuring semantic similarities by combining gene ontology annotations and gene co-function networks

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Gene Ontology (GO) is widely used for studying gene functional relationships.
• Current semantic similarity measures rely solely on GO annotations and structure.
• Limited gene annotation data in many organisms restricts the power of existing GO-based similarity measures.

Purpose of the Study:

• To introduce a novel network-based similarity measure (NETSIM) for improved gene functional analysis.
• To enhance the accuracy of GO term similarities by incorporating gene co-function networks.
• To provide a method that performs well even with sparse gene annotation data.

Main Methods:

• Developed NETSIM, a novel approach integrating GO annotations, GO structure, and gene co-function networks.
• Utilized metabolic reaction maps from yeast, Arabidopsis, and human for validation.
• Applied NETSIM to large gene families, such as cytochrome P450 monooxygenases in Arabidopsis.

Main Results:

• NETSIM significantly improves the accuracy of GO term similarities.
• The method demonstrates effectiveness even for genomes with sparse gene annotation data.
• Functional grouping of Arabidopsis cytochrome P450 monooxygenases facilitated their characterization.

Conclusions:

• Incorporating genome-specific information, like co-function networks, substantially improves semantic similarity measures.
• NETSIM leverages both GO data and gene co-function networks as prior knowledge.
• This enables the measurement of taxon-specific functional similarities not explicit in GO, especially when annotations are limited.