Dec. 17, 2007 | According to MITRE's Lynette Hirschman, the top performing entries for the gene mention task at BioCreative II included an entry from IBM, led by Rie Ando Johnson, and two systems from Taipei: National Yang-Ming University and Academia Sinica. All three systems used advanced machine learning techniques.
The best performances in the gene normalization task all came from Germany: Jörg Häkenberg (Technical University, Dresden), Katrin Fundel (Ludwig-Maximilian University, Munich), and Juliane Fluck (Fraunhofer Institute). Häkenberg's system used synonyms and contextual information to associate gene identifiers with mentions in text. The other groups used large-scale synonym lexicons coupled with string matching techniques.
Among the top groups in the protein-protein interaction annotation task were Larry Hunter's lab (Univ. Colorado School of Medicine) and Claire Grover's group (University of Edinburgh). Both combined machine learning with manually derived rules to create pipelines to flag articles for curation, find relevant genes and proteins, and identify specific interacting proteins.
Return to main article.