Bio-Strings: A Relational Database Data-Type for Dealing with Large Biosequences

Abstract

DNA sequencers output a large set of very long biological data strings that we should persist in databases rather than basic text file systems. Many different data models and database management systems (DBMS) may deal with both storage and efficiency issues regarding genomic datasets. Specifically, there is a need for handling strings with variable sizes while keeping their biological meaning. Relational database management systems (RDBMS) provide several data types that could be further explored for the genomics context. Besides, they enforce integrity, consistency, and enable good abstractions for more conventional data. We propose the relational text data type to represent and manipulate biological sequences and their derivatives. We present a logical schema for representing the core biological information, which may be inferred from a given biological conceptual data schema and the corresponding function manipulations. We implement and evaluate these stored functions into an actual RDBMS for both efficacy and efficiency. We show that it is possible to enforce basic and complex requirements for the genomic domain. We claim that the well-established relational text data type in RDBMS may appropriately handle the representation and persistency of biological sequences. We base our approach on the idea of domain-specific abstract data types that can store data with semantically defined functions while hiding those details from non-technical end-users.