Searching for geneRIFs: Concept-Based Query Expansion and Bayes Classification

k gene annotations. Our indexing machinery produces per indexed MEDLINE abstract a list of concepts with an accompanying weight, termed a fingerprint. Searching is done by matching a query int against fingerprints of all indexed MEDLINE abstracts. Query fingerprints are generated by combining fingerprints of four types. First, a fingerprint containing just the gene concept with all the ne names and aliases. Second, a combination of MEDLINE fingerprints of all abstracts in which the gene concept was found without ambiguity problems. Third, a generic fingerprint with concepts typical of geneRIFs, when compared to MEDLINE in general. Fourth, a fingerprint containing the concepts of the Gene Ontology (GO) annotation When it comes to identifying a gene name in a text the large number of synonyms and the frequent occurrence of homonymy are problematic. In our approach we attempt to deal with both. Synonymy as found in Locuslink is incorporated in our thesaurus. An attempt was made to reduce ...

• Type: Conference
• Year: 2003
• Where: TREC
• Authors: Rob Jelier, Martijn J. Schuemie, C. Christiaan van der Eijk, Marc Weeber, Erik M. van Mulligen, Bob J. A. Schijvenaars, Barend Mons, Jan A. Kors
• Keywords: Gene, Gene Concept, Indexed Medline Abstracts, TREC 2003, TREC 2008