A multi-infrastructure gateway for virtual drug screening

In computer-aided drug design, software tools are used to narrow down possible drug candidates, thereby reducing the amount of expensive in vitro research, by a process called virtual screening. This process includes large computations that require advanced computing infrastructure; however, using rapidly evolving high-performance computing platforms can be difficult for biochemists. In this paper, we present a science gateway for virtual screening that has been tailored to the specific needs of our local users. The gateway provides user-friendly access to distributed computing infrastructures for high-throughput experiments with a few clicks. Its design is based on the generic layer developed for another gateway for neuroimaging data analysis, including data and computation management, as well as support for its operation. To facilitate scalability, the system architecture allows for adding new computing platforms to the back-end without affecting the front-end, from which the user can dynamically choose the preferred infrastructure. This paper describes the user-centered design process, the system architecture and a performance assessment using gLite grid, Hadoop, and a local cluster. The empirical results show considerable speed-ups and ease of use, as well as user satisfaction. Copyright (c) 2015John Wiley & Sons, Ltd.