IntelliGO: a new vector-based semantic similarity measure including annotation origin.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Gene Ontology (GO) is a crucial controlled vocabulary for gene annotation.
• Existing semantic similarity measures often do not account for GO annotation origins (evidence codes).
• There is a need for more comprehensive gene similarity measures that consider annotation evidence.

Purpose of the Study:

• To introduce IntelliGO, a novel semantic similarity measure for gene annotation.
• To develop a method that integrates information content and evidence codes within a vector space model.
• To evaluate IntelliGO's performance against existing measures using benchmark datasets.

Main Methods:

• Developed IntelliGO using a vector space model integrating GO term information content and evidence code weights.
• Adapted generalized cosine similarity for the GO graph structure.
• Tested IntelliGO on KEGG pathways and Pfam domains for yeast and human genes, involving over 67,900 comparisons.

Main Results:

• IntelliGO demonstrated superior performance in expressing biological cohesion of gene sets compared to four existing measures.
• The measure effectively discriminated between distinct gene sets in inter-set comparisons.
• IntelliGO allowed for the evaluation of evidence code weight influence and showed good correlation with other classification methods.

Conclusions:

• IntelliGO offers a customizable and comprehensive approach to quantifying gene similarity using GO annotations.
• Its robust set-discriminating power makes it suitable for functional clustering applications.
• The measure provides valuable insights into gene relationships by considering annotation evidence.