How Can Semantic Annotations Support the Identification of Network Similarities?

Computational models in open model repositories support biologists in understanding and investigating biological questions. The availability of alternative models results in a need for model selection algorithms. This selection can be based on information retrieval search, full-text search, selection by ontology concepts etc. We here describe our approach to solving a serious aspect of model selection, that is, the problem of comparing models of reaction networks. Specifically, we discuss how graph algorithmic approaches can help to compare models that are semantically enriched by annotations. While a graph comparison of naked models is infeasible, the knowledge gained from the semantic annotations and domain specific structures can reduce the complexity. Our concept has the potential to improve model search, and it can contribute to the definition of similarity measures.