MoBIoS Index: Support Distance-Based Queries in Bioinformatics

Given a database, answering a distance-based query means retrieving all the data objects in the database that are in close proximity to a query object. Proximity is defined by any metric distance function. Close can mean within a certain distance, a range query, k-nearest neighbor or the two in combination. Answering distance-based queries is a fundamental component of many biological applications, including sequence homology, protein identification by spectral database look-up and biomedical image retrieval. Most systems for retrieving similar biological data objects are domain-specific. For each new model of similarity the retrieval problem must be revisited. (Metric) distance-based indexing only requires that the data can be abstracted into metric space, enabling the reuse of the same software package for many problems. The Molecular Biological Information System (MoBIoS) is a metric-space DBMS targeting bioinformatics applications. In this paper we describe the programmatic interface for MoBIoS index methods. In addition to built-in metrics, the interfaces enable users to integrate new metrics. The system supports four different retrieval methods. We characterize these methods and their applicability to different problems.