High Performance Computing Initiative

The biomedical sciences have made a tremendous progress in the past decade as a result of high-throughput instrumentation in the field of genomics and related high-performance computational data analysis and management. This trend continues as current biomedical research has been driven by the urgent need to translate those biological discoveries into the clinical practice to improve human health. Processing and analyzing gigabytes, even terabytes, of data routinely at reasonably responsive intervals are increasingly critical challenges at the forefront of biomedical research and education. This is especially true in the fields of high-throughput genomics, proteomics, and neuro-imaging.

To address these critical challenges, the Department of Biostatistics, Bioinformatics and Epidemiology (DBBE) at the Medical University of South Carolina seeks to initiate a high-performance computing project for computation and data intensive research across a broad spectrum of disciplines - more specifically the use of parallel computing infrastructure to enhance discovery and education efforts in the biosciences. By leveraging the use of high performance systems it becomes possible to empower researchers and engage the MUSC community in a collaborative approach to emerging technologies.

This project serve to provide the foundation for teams of researchers and educators to act upon high volume data through a collaborative approach rather than in isolated environments. Additionally, this high performance computing initiative will foster new discovery through data driven experiments and seek multi-disciplinary approaches to solve problems in genomic and proteomic research, predictive modeling and clinical studies.

A prototype system "BioMed Engine" was developed by the DBBE to test the feasibility and usability of cluster-based high performance computing. It is a Linux cluster built to provide a high performance, scalable, parallel computing environment for biomedical related research. The BioMed Engine was built with a HP ProLiant DL380 G3 server, HP ProLiant Blade G2 servers, ProLiant BL p-Class GbE Interconnect Switches, Tera Byte Storage, and a cluster distribution called Rocks. Rocks is a complete cluster package based on Redhat Linux with the latest kernel and management utilities. The results were exceptional in the fields of study including Genomics, Proteomics, Neuro-Imaging, and Bioinformatics. Web front-ends were also developed to facilitate use of HPC tools by the biomedical researchers by providing a user-friendly and uniform interface, which hides the technical details of high-performance computing.

A set of tools (MatlabMPI, mpiBLAST, mpiBLAST-PIO, SNOW, OMSSA, PBS, SGE etc.) enabling parallel or distributed computation for biomedical research have been implemented on the BioMed Engine. The active project involves developing parallel bayesian probabilistic component analysis algorithms on fMRI deception data.

The HPC Initiative is actively looking for collaborators in the fields of cell biology, biochemistry, genomics, proteomics, neuroscience and bioengineering etc. to explore its computational and storage capacity. It is also looking for hardware grants to further expand its computational facility.

The web portal for BioMed Engine is http://hpcc.musc.edu.