The full text of this thesis is available from the Newcastle University Library website: https://theses.ncl.ac.uk/dspace/handle/10443/1410
Flanagan, K.S., School of Computing Science, University of Newcastle upon Tyne
Recent advances in genome sequencing technologies have unleashed a flood of new data. As a result, the computational analysis of bioinformatics data sets has been rapidly moving from a lab-based desktop computer environment to exhaustive analyses performed by large dedicated computing resources. Traditionally, large computational problems have been performed on dedicated clusters of high performance machines that are typically local to, and owned by, a particular institution. The current trend in Grid computing has seen institutions pooling their computational resources in order to offload excess computational work to remote locations during busy periods. In the last year or so, commercial Cloud computing initiatives have matured enough to offer a viable remote source of reliable computational power. Collections of idle desktop computers have also been used as a source of computational power in the form of 'volunteer Grids'. The field of bioinformatics is highly dynamic, with new or updated versions of software tools and databases continually being developed. Several different tools and datasets must often be combined into a coherent, automated workflow or pipeline. While existing solutions are available for constructing workflows, there is a clear need for long-lived analyses consisting of many interconnected steps to be able to migrate among Grid and cloud computational resources dynamically. This project involved research into the principles underlying the design and architecture of flexible, high-throughput bioinformatics processes. Following extensive research into requirements gathering, a novel Grid-based platform, Microbase, has been implemented that is based on service-oriented architectures and peer-to-peer data transfer technology. This platform has been shown to be amenable to utilising a wide range of hardware from commodity desktop computers, to high-performance cloud infrastructure. The system has been shown to drastically reduce the bandwidth requirements of bioinformatics data distribution, and therefore reduces both the financial and computational costs associated with cloud computing. The system is inherently modular in nature, comprising a service based notification system, a data storage system scheduler and a job manager. In keeping with e-Science principles, each module can operate in physical isolation from each other, distributed within an intranet or Internet. Moreover, since each module is loosely coupled via Web services, modules have the potential to be used in combination with external service oriented components or in isolation as part of another system. In order to demonstrate the utility of such an open source system to the bioinformatics community, a pipeline of inter-connected bioinformatics applications was developed using the Microbase system to form a high throughput application for the comparative and visual analysis of microbial genomes. This application, Automated Genome Analyser (AGA) has been developed to operate without user interaction. AGA exposes its results via Web-services which can be used by further analytical stages within Microbase, by external computational resources via a Web service interface or which can be queried by users via an interactive genome browser. In addition to providing the necessary infrastructure for scalable Grid applications, a modular development framework has been provided, which simplifies the process of writing Grid applications. Microbase has been adopted by a number of projects ranging from comparative genomics to synthetic biology simulations.