Biomed at PSC
A Computational Environment for Biomedicine
by David Deerfield, Michael Levine, Ralph Roskies & Joel Stiles
High-performance computing and communications (HPCC) has become increasingly central to leading-edge biomedical research. In the last decades, biomedicine has changed from a mostly data-poor, qualitative science to one that is increasingly data-rich, increasingly quantitative, and heavily influenced by high-throughput technologies. HPCC itself has also changed dramatically, from a world dominated by large purpose-built mainframes to much more powerful systems that are increasingly distributed, and are increasingly affordable at modest sizes. Adequate access to HPCC resources inspires scientists to attack problems that would otherwise be considered infeasible. High-performance computing, along with the expertise needed to utilize it, redefines the kinds of problems that scientists are willing to investigate.
The PSC Biomedical Research Group The PSC biomedical group comprises 13 full-time staff, including seven PhD scientists, five master’s degree programmer-analysts in computer science and engineering and a full-time administrator, along with 10 interns and four students.
Over the past several years, leadership at the National Institutes of Health (NIH) has put a premium on HPCC and its critical role in future breakthrough advances for human health. This critical role is reflected in the NIH’s recent Roadmaps, and is often motivated by the continuing explosion in range and volume of biomedical data. However, biological understanding also requires advances in modeling and simulation that will require HPCC on previously unforeseen scales.
In recent testimony before the President’s Information Technology Advisory Council (PITAC), Eric Jakobsson defined a mission statement for the NIH’s Roadmap on Bioinformatics and Computational Biology. Chair of NIH’s Bioinformation Science and Technology Initiative Consortium and director of the National Institute of General Medical Sciences Center of Bioinformatics and Computational Biology, Jakobsson stated:
In ten years, we want every person involved in the biomedical enterprise — basic researcher, clinical researcher, practitioner, student, teacher, policy maker — to have at their fingertips through their keyboard, instant access to all the data sources, analysis tools, modeling tools, visualization tools, and interpretative materials necessary to do their jobs, with no inefficiencies in computation or information technology being a rate-limiting step.
While a ten-year timescale may be overly ambitious, on a longer timescale, we at PSC can envision a groundbreaking “Computational Environment for Biomedicine” (CEB) that will link all Genomics, Proteomics, and other growing databases with integrated tools for imaging, model generation, multiscale simulation and visualization, analysis, training, education, and HPCC- and grid-based computing. In addition, such a CEB will be linked to increasingly personalized information on genomic sequence, time-dependent protein expression levels, and many other quantitative measures obtained on a case-by-case basis. All of this population and individualized data coupled to a CEB will enable breakthrough clinical advances such as:
- predicting the whole body effect of a genomic mutation;
- drug design, including prediction of systemic side effects;
- halting and reversing neurodegenerative changes;
- anti-aging therapies;
- bioengineering and materials science solutions to sensory deficits, cardiovascular disease, and a host of musculoskeletal defects;
- understanding and intervening effectively in psychiatric disorders.
Of course, this will not be an easy task. In his PITAC presentation, Jakobsson also pointed out that many of today's limitations in biomedical computing stem from software that:
- is difficult to use,
- is fragile,
- lacks interoperability of different components
and further pointed out the ongoing shortage of personnel trained to create and use better biological computing tools and environments.
To a large extent, these shortcomings have been addressed by the PSC’s Research Resource since its inception in 1987, when the PSC became the first extramural supercomputing center to receive funds from the NIH. Our goal for the future remains invariant — to develop and use HPCC to benefit biomedical research. The PSC Research Resource brings together experts from multiple disciplines, and provides access to cutting-edge computing by leveraging investments of other agencies. It has made software easier to use, made it more robust, presented it as part of an integrated framework, and has continually emphasized training and outreach. With the flexibility to respond quickly to unanticipated research opportunities, it has expanded the research domains that can effectively use HPCC from structural biology and sequence-based bioinformatics to include neural modeling, pathology, intra- and intercellular modeling, and visualization and analysis of extremely large imaging datasets, all components for a CEB.
We believe that the next step critical to realizing the vision of a CEB and the benefits it will bring to human health lies in realistic cell-to-organism level modeling. It is strategically positioned at an overlapping region of Structural Molecular Biology, Bioinformatics, Imaging, and Systems Biology, and is also strategically positioned between “bottom-up” and “top-down” approaches to biochemical network identification and functional analysis. Realistic physiological simulations present a grand challenge because of the wide range of underlying space and time scales, as well as the widely disparate organization and properties of different cells. Spatially realistic modeling will require new multiscale algorithms and prodigious amounts of computing, databases, storage, networking, and visualization. Such prodigious effort and costs, however, will be overwhelmingly outweighed by the resulting benefits to individuals and society at large. The PSC Research Resource will be a national leader in realizing these goals.