Evaluation of Protein-Protein Interaction Management Systems

Abstract

Protein-protein interactions (PPIs) are very important for observing the behavior of known proteins in biological processes and in the study of many diseases. Currently, there is a number of PPI databases publicly available in the Web. In most cases, these datasets are managed by traditional relational database management systems (RDBMS) and they are shared as plain or XML files. A very useful approach would be the the unification of these separated data sources following the Semantic Web linked open data (LOD) paradigm, in order to complement and extend the existing knowledge of each data source. Semantic representation and storage of linked open datasets can be performed by many off-the-shelf systems modeling them as Resource Description Framework (RDF) models. RDF modeling, provides great flexibility for the linking, querying and mining of various PPI data sources. In this paper, we evaluate experimentally the interconnection and storage of various PPI data sources with off-the-shelf RDF storages. We examine the performance of such storages against traditional RDBMS in the context of PPI dataset management. Our main findings show that each one of the alternative storage methods, has its own advantages and disadvantages (in processing time and memory utilization) according to various types of queries.