by Luella LeVee
One of the brightest lights emerging from DCRT's recent reorganization is the Computational Bioscience and Engineering Laboratory (CBEL), whose major thrust, according to CBEL Chief Robert Martino, is "exploiting high-performance computer systems for biomedical applications." Much of the work of this new laboratory centers on a highly parallel supercomputer, the Intel iPSC/860, whose claim to fame is a fantastic computational speed that has helped more than 50 NIH scientists get research results in minutes instead of hours.
"It's exciting and a continuing challenge to have high-performance computing impact biomedical research and clinical practice, to observe applications in daily practice," says Martino.
CBEL's expertise in highly parallel computing is already speeding up some of NIH's work in image processing, structural biology, computational chemistry, and medical imaging. In addition, CBEL staff are providing NIH researchers with faster ways to visualize biological processes, search genetic databases, and conduct linkage and statistical analyses. "With our help, scientists are getting the latest up-to-date computing technology both in hardware and software and the best expertise to help them solve their problems using this technology," says Martino.
The Benefits of Parallel Computing
Until recently, these scientific tasks were accomplished with conventional computers, where problems are solved sequentially, in a step-by-step fashion. In parallel computing, on the other hand, a problem is divided into several segments and each segment is sent to a different processor or node. The segments are then computed simultaneously, saving precious time for scientists.
"Everyday-workstation users can now access the parallel computer over the NIH network. For example, users can now access the parallel computer at their workstation to complete in less than 5 minutes, an imaging task that marmally takes 6 hours on a workstation," says Martino. CBEL already has a few success stories to tell:
* With CBEL's help, NIAMS collaborators have a better understanding of the structure of the herpes simplex virus type 1. Parallel computing helped to determine the location of the major types of proteins that combine to form the virus's capsid.
* NIDDK has used a parallel-computing method to automate the spectral-assignment process in NMR spectroscopy, determining which signals in the multidimensional NMR spectra data belong to which atoms in the molecule under study. Using this method, the scientists assess the structure of calmodulin, a protein involved in a wide range of cellular-calcium-dependent signaling pathways.
* Another group of scientists from NIDDK used the parallel computer to simulate the kinetics of ultrafast chemical reactions in solution, such as the kinetics of nitric oxide rebinding to myoglobin following photodissociation. The method yielded insights into the chemical dynamics of ligand binding to myoglobin.
* NIMH investigators used parallel-image-registration techniques developed by CBEL staff and positron emission tomography (PET) images of the brain superimposed on computer tomography (CT) and magnetic resonance imaging (MRI) images to study the progression of Alzheimer's disease.
* High-performance computing has allowed NEI researchers to determine the onset time, the rate of information encoding, and the total amount of information encoded by neuronal responses to a visual stimulus in primates. This will help researchers to develop better models of the primate visual system.
"The most exciting part of my work is bringing technology to a significant biomedical problem. I really enjoy seeing the effect it has on biomedical applications," says Martino.
CBEL is also participating in Vice President Gore's High Performance Computing and Communications Initiative, an interagency program to bring high-performance computing to bear on "grand challenge" problems, such as predicting protein folding and designing drugs, and "national priorities," such as health-care reform.
CBEL's help is available to all NIH scientists, says Martino. "Our goal is not only to help scientists with computationally intensive problems taking a very long time on existing systems, but it's also getting scientists to think of new ways of approaching their problems that they haven't considered before because they didn't have the computational power. Sometimes, it is the software they are using, and we can assist through software engineering. Other times, a problem is appropriate for a parallel computer and we are able to help them out that way."
Dr. Martino invites NIH scientists with computationally intensive problems to call him at 496-1111.