A New Database for Drug Discovery Through Application of Data-Integration and Semantics

Abstract

Biomedical databases are important in drug- discovery. Example applications are: disease-treatments, investi- gating side-effects of drugs, and identification of similar research- efforts. A challenge in construction of biomedical databases concerns the complexities they describe: bio-medical data are collected from a large number of data-resources, where each of the data-resources are targeted towards a specific audience. What we observe is that the current biomedical databases for semantic searches does not address the needs of drug- discovery. Drug-discovery require data from multiple sources to be meaningfully unified/glued into a semantic database, ie, to provide/give correct results to complex queries. In practice established approaches ignore the latter. Implication is that biomedical semantic databases are not used in drug-discovery. In this paper we address the latter issues. We present a novel database which combines strategies for model-unification of 37 external databases, unwrapping of BioPax formatted data-sets, data-normalization, and semantic inferences. From our empirical evaluation we observe how our approach captures the effects of causation in bio-medical pathways. Therefore our biomedical database provides support for accurate drug-discovery. Our new semantic database is freely accessible through a user-friendly semantic search-interface at www.knittingTools.org.