M-Finder: Uncovering functionally associated proteins from interactome data integrated with GO annotations

Table 1 Summary of semantic similarity methods in four categories.

Method	Description
Edge-based
Path-length	Path-length between two GO terms
Depth	Depth to SCA divided by average depth to two GO terms
Node-based
TO	The number of common ancestors of two GO terms
simUI	Common ancestors divided by union of ancestor sets of two GO terms
Annotation-based
Resnik	IC of SCA of two GO terms
Lin	IC of SCA divided by average IC of two GO terms
Jiang	Sum of differences of ICs between SCA and two GO terms
GraSM	Average IC of all disjunctive common ancestors of two GO terms
simRel	Combination of Resnik's and Lin's methods
simICND	Combination of Resnik's and Jiang's methods
Integrative
G-SESAME	Combination of common ancestor terms and their depth
simGIC	Combination of simUI and ICs of ancestor terms
IntelliGO	Combination of depth to two GO terms and ICs of ancestor terms
TCSS	Combination of Resnik's method and a clustering technique
simICNP	Combination of Resnik's method and path-length between two GO terms

We group the existing semantic similarity methods and two proposed measures (simICND and simICNP) into four broad categories according to the components used in GO. SCA denotes the most specific common ancestor term of two GO terms that have the annotation of two proteins of interest, respectively. IC denotes the information content of a GO term.

ISSN: 1477-5956