IDPredictor : predict database links in biomedical database

Abstract

Knowledge found in biomedical databases, in particular in Web information systems, isa major bioinformatics resource. In general, this biological knowledge is worldwide rep-resented in a network of databases. These data is spread among thousands of databases,which overlap in content, but differ substantially with respect to content detail, interface,formats and data structure.To support a functional annotation of lab data, such as protein sequences, metabolites orDNA sequences as well as a semi-automated data exploration in information retrieval en-vironments, an integrated view to databases is essential. Search engines have the potentialof assisting in data retrieval from these structured sources, but fall short of providing acomprehensive knowledge excerpt out of the interlinked databases. A prerequisit of sup-porting the concept of an integrated data view is to acquire insights into cross-referencesamong database entities. This issue is being hampered by the fact, that only a fraction of allpossible cross-references are explicitely tagged in the particular biomedical informationssystems.In this work, we investigate to what extend an automated construction of an integrated datanetwork is possible. We propose a method that predicts and extracts cross-references frommultiple life science databases and possible referenced data targets. We study the retrievalquality of our method and report on first, promising results.