Mapping gene/protein names in free text to biomedical databases

Observing that many biomedical databases have been developed and maintained independently, their records referring to the same entities may have different sets of synonyms. Integration of names pertaining to the same entity would provide a more comprehensive list of synonyms than each individual database. We have assembled BioThesaurus, a thesaurus of proteins and their corresponding genes compiled from multiple databases for all UniProtKB records. In this study, the coverage of BioThesaurus, and the contribution of each individual database were assessed for several organisms. The result indicates that the coverage of BioThesaurus is over 80% for most of the organisms with an average of 85.4%. When restricted to individual databases or resources, the percentages dropped ranging from 3 to 30%. The study demonstrated that each individual database or resource has some synonyms not covered by other databases or resources, and a list of names compiled from multiple databases would be desired for systems requiring high recall.