Major advances and an explosion of new information in the biomedical sciences are in part the result of our ability to pose and solve questions of biological structure and function at all levels and on all scales, thanks to advanced computation. Today a host of critical problems depend on large-scale computation and scientific visualization. The most advanced machines of the hour are massively parallel supercomputers. We propose, therefore, to create an NIH National Biomedical Computation Resource (NBCR) to accomplish three major purposes: (1) to carry out important biomedical research programs that will benefit from and can extend the new technologies for computation and visualization, (2) to assemble a group of core programs that focus on different scales of biological activity, with a thrust toward developing ways of accessing interactions among and between scales, and (3) to use those programs as drivers to make the new technologies friendly and accessible to a larger community of biomedical researchers. NBCR will give biomedical researchers access to a 400-processor Intel Paragon supercomputer at the San Diego Supercomputer Center (SDSC). SDSCs broad program of research in parallel computing emphasizes the development of operating systems, applications, and advanced visualization. Thus a high level of support can be mobilized to aid scientists using the parallel hardware for modeling, simulation, analysis, and visualization. Core and collaborative research at NBCR will address several important and often closely related topics. The improvement of methods for calculating density functional integrals will speed the process of obtaining geometries and force fields for large molecules of biological interest and will result in a DFT package on the Paragon for use by NBCR researchers. The concurrent parallel implementation of the AMBER molecular dynamics package will enable researchers to combine the virtues of quantum mechanical and molecular dynamics programs. Improved methods of analyzing patterns in protein sequences and structures will be developed for use in genomic studies and modeling homologous protein structures. Visualization packages will be developed as part of and in conjunction with these programs. In addition, work will be carried out on the prediction of RNA structure at kilobase lengths, sizes appropriate to studying gene expression in RNA viruses (e.g., HIV), RNA processing, and phylogenetics; and on finite-element modeling of the electromechanics of cardiac muscle, as a paradigm computational tissue-function study. NBCR will fund a portion of the processor complement of the Paragon, guaranteeing the allocation of a minimum of 140,000 processor hours annually to NBCR researchers. Enhancements to the capabilities of SDSCs VisLab will support NBCR advances in imaging. NBCR staff will port existing applications to the Paragon and work with core and collaborative investigators on new or revised applications. Workshops and courses will be given to train all NBCR researchers to make best use of the resource, and resource information and research results will be disseminated to users and the wider biomedical community.
Showing the most recent 10 out of 270 publications
