A new method based on semantic similarity between GO terms to predict gene functions

With the increasing amount of detected genes and gene products, more and more unknown functions of genes and gene products need to be predicted. A lot of methods are proposed to predict gene functions, but most of them neglect the existing annotation information. This article proposes a novel method to predict function similarity between genes and gene products based on existing annotation information using probabilistic derivation. We first calculate the Gene Ontology (GO) term´s semantic similarity by gathering the influence of their ancestor terms in its directed acyclic graph. Then, the similarity of GO terms is obtained to calculate the function similarity of genes by a designed algorithm. The result of the experiments indicates that the algorithm has a good performance in predicting the function similarity between genes and gene products.