MetReS: a Metabolic Reconstruction Database for Cloud Computing

When designing a cloud infrastructure, it is critical to ensure beforehand that the system will be able to offer the desired level of QoS (Quality of Service). Our attention is focused here on efficient QoS accessing to a biological database in cloud computing systems. Our group developed two software applications that address important biological problems, Biblio-MetReS and Homol-MetReS. Biblio-MetReS is a data-mining tool that facilitates the reconstruction of molecular networks based on automated text-mining analysis of published scientific literature. Homol-MetReS allows functional (re) annotation of proteomes, to properly identify both the individual proteins involved in the process(es) of interest and their function. Reconstruction of molecular networks is essential to understand how organisms work at the molecular level and has strong implication, for example, in finding targets to treat different types of disease. In addition, the identification and functional annotation of the individual components of the network is crucial to understand what those targets might do in the context of the organism. These two software applications access the same database of organisms with annotated genes. The efficiency of the two applications is directly related to the design of the shared database. This database is continuously growing, as hundreds to thousands of new genomes are sequenced and annotated each year. The main goal of the current work was to improve the current database performance and to test if this improvement would scale to larger data-sets and more complex types of analysis that are not yet done by either of the applications. To achieve this goal, different database architectures were designed and analyzed. We started the study with a public relational database, MySQL, which was the current database server used by these applications. Then, due to the large size of the database, Apache Hadoop, a framework used for large-scale data processing, was considered and studied as an alternative. Although Big Data systems are not always a replacement of traditional relational databases, we proved by extensive tests the applicability of Apache Hadoop to a standard biological database containing some of the most frequently used types of information in molecular and systems biology. With time, as this database will continuously grow, the proposed solution will further improve its efficiency. Furthermore, this solution allows to extract additional valuable information from the data-sets that was not being currently considered.