As the Gene Ontology (GO) plays more and more important roles in bioinformatics research, there has been great interest in developing objective and accurate methods for calculating semantic similarity between GO terms. In this chapter, the authors first introduce the basic concepts related to the GO and then briefly review the current advances and challenges in the development of methods for calculating semantic similarity between GO terms. Then, the authors introduce a semantic similarity method that does not rely on external data sources. Using this method as an example, the authors show how different properties of the GO can be explored to calculate semantic similarities between pairs of GO terms. The authors conclude the chapter by presenting some thoughts on the directions for future research in this field.
TopSemantic Similarity Between Gene Ontology Terms
The calculation of semantic similarity between pairs of ontology terms aims to capture the relatedness between the semantic content of the terms. Researchers have made great efforts to develop objective and accurate methods to calculate term semantic similarity. For example, semantic similarity between concepts has been a central topic in natural language processing where several robust methods have been proposed based on the WordNet ontology (Budanitsky & Hirst, 2006). In recent years, ontologies have grown to be a popular topic in the biomedical research community creating a demand for computational methods that can exploit their hierarchical structure, in particular, methods for calculating semantic similarity between terms in the GO. Such methods are designed to reflect the closeness or distance between the semantic content of the terms, in other words, their biological relationships.
Additionally, semantic similarity methods can easily be extended to infer higher level semantic relationships. For example, at the protein level, scores for a given protein pair can be calculated by combining the pairwise semantic similarities for the GO terms associated with the proteins. These scores can be used in a broad range of applications such as clustering of genes in pathways (Wang, Du, Payattakool, Yu, & Chen, 2007, Sheehan, Quigley, Gaudin, & Dobson, 2008, Nagar & Al-Mubaid, 2008, Du, Li, Chen, Yu, & Wang, 2009), protein-protein interaction (Xu, Du, & Zhou, 2008), expression profiles of gene products (Sevilla et al., 2005), protein sequence similarity (Pesquita et al., 2008, Mistry & Pavlidis, 2008, Lord, Stevens, Brass, & Goble, 2003), protein function prediction (Fontana, Cestaro, Velasco, Formentin, & Toppo, 2009), and protein family similarity (Couto, Silva, & Coutinho, 2007). An armada of semantic similarity measures using the GO are available in the biomedical literature. A representative collection of available methods have been reviewed and categorized by (Pesquita, Faria, Falcão, Lord, & Couto, 2009).