Determining the semantic similarities of GO terms based on their existence dependencies

We introduce in this paper a system called EDGT, which determines the semantic relationships among Gene Ontology terms. EDGT accepts Keyword-based queries with the form Q (“t₁”, “t₂”, .., “t_n”) and Loosely Structured queries with the form Q( “t₁”,..., “t_n”, R₁?,..., R_n?), where t_i denotes a GO search term and R_i denotes a return/result GO term. The result of the Keyword-based query Q(“t₁”, “t₂”) is a set of GO terms, where each term in the set is semantically related to both t₁ and t₂. The framework of EDGT determines the relationships among GO terms using the concept of existence dependency. EDGT resides between user application and Gene Ontology database. We experimentally evaluated EDGT and compared it with four existing methods. The results showed marked improvement.