GO Semantic Similarity Based Analysis for Huaman Protein Interactions

Background: Because of the importance of protein interaction in organisms, researchers interested in the functional similarity between interacted proteins. Gene Ontology semantic similarity provide a novel way to measure the similarity between gene products. Various methods have been proposed for calculating the semantic similarity between GO terms and gene products, therefore it is a serious problem to choose appropriate approach for specific application. Result: Based on a comparible smaller random selected dataset, largescale computational experiments have performed to evaluate 24 different definitions of GO semantic similarity between human proteins. In order to investigate the different effective of different calculating methods, both of the best and the worst performed definitions have been selected to analysis whole human protein interaction network. Conclusion: GO semantics similarity is a effective tool to analysis human protein interactions, but choosing a appropriate approach is key step in such work. For human protein interaction analysis, mean is appropriate for calculating the similarity between gene products and different methods are appropriate for different categories of GO terms. Website: Support information and related source codes of this paper can be found on http://bioinfo.csu.edu.cn/chengang/goana/.