Statistical search on the semantic web

Motivation: Statistical analysis of links on the Semantic Web is important for various evaluation purposes such as quantifying an individuals scientific research output based on citation links. SPARQL has been proposed as a standardized query language for the Semantic Web and is intuitively understandable; however, it does not adequately support statistical evaluation of semantic links. Results: We have extended SPARQL to a novel Resource Description Framework (RDF) query language termed General and Rapid Association Study Query Language (GRASQL) to generate inferences connecting semantic Boolean-based deduction and statistical evaluation of RDF resources. We have verified the descriptive capability of GRASQL by writing GRASQL queries for practical biomedical search patterns including in silico positional cloning studies and for ranking researchers in a specific domain of expertise by introducing k index, the number of papers containing specific keywords that are published in a fixed period by a researcher. We have also developed a search engine termed General and Rapid Association Study Engine (GRASE), which executes a restricted variety of GRASQL queries by requesting a dynamic and comprehensive evaluation of statistical significance of intersections between each group of documents assigned to URIs and those documents matching user-specified keywords and omics conditions. By performing practical in silico positional cloning searches with GRASE, we show the relevance of our approach on the Semantic Web for biomedical knowledge discovery problem solving.