Ontology-based Similarity Measures and their Application in Bioinformatics

Genome-wide sequencing projects of many different organisms produce large numbers of sequences that are functionally characterized using experimental and bioinformatics methods. Following the development of the first bio-ontologies, knowledge of the functions of genes and proteins is increasingly made available in a standardized format. This allows for devising approaches that directly exploit functional information using semantic and functional similarity measures. This thesis addresses different aspects of the development and application of such similarity measures. First, we analyze semantic and functional similarity measures and apply them for investigating the functional space in different taxa. Second, a new software program and a new database are described, which overcome limitations of existing tools and simplify the utilization of similarity measures for different applications. Third, we delineate two applications of our functional similarity measures. We utilize them for analyzing domain and protein interaction datasets and derive thresholds for grouping predicted domain interactions into low- and high-confidence subsets. We also present the new MedSim method for prioritization of candidate disease genes, which is based on the observation that genes and proteins contributing to similar diseases are functionally related. We demonstrate that the MedSim method performs at least as well as more complex state-of-the-art methods and significantly outperforms current methods that also utilize functional annotation