In Progress, 2009

Bending in Cellular Skeletons

Researchers use BigBen to better understand the structural engineering of protein filaments that maintain the shape of cells

An actin trimer (three-part molecule), a fundamental building block in the cytoskeletal filament, with bound phalloidin (red). Amino-acid residues in contact with phalloidin are shown with red spheres

How do cells maintain their shape? What gives them structure? Similar to the skeleton of the body, the “cytoskeleton” of cells maintains cellular shape, protects the cell, and plays important roles in cellular function. One of the main components of the cytoskeleton in eukaryotic cells — cells with membranes, such as human cells — are filaments formed from chains of a protein called actin.

Actin chains form a highly branched network, and though much is understood, many fundamental questions remain about actin filaments and about how this network forms and functions. Researchers led by University of Utah chemist Greg Voth used PSC's BigBen and systems at other TeraGrid sites to simulate the actin cytoskeleton, and their findings are beginning to provide answers to some of these questions.

Using 512 BigBen processors for several weeks, Voth and colleagues investigated the structure and dynamics of two competing actin filament models. “Our simulations of the actin filament performed at PSC have shown us, for example, that the individual actin monomers in the actin filament undergo a large conformational change on relatively short timescale,” said Voth collaborator James Pfaendtner, a biomolecular engineer at the University of Washington.

For this work, PSC scientist Phil Blood optimized performance of the latest version of the molecular dynamics program NAMD. Results have also helped the team to shed new light on how the formation of actin chains is involved in ATP hydrolysis — a key issue in cytoskeletal dynamics. The researchers also carried out the first simulations of the actin filament bound with phalloidin — a toxic compound widely used to assist researchers in creating images of the cell, and their results have begun to clarify the atom-by-atom mechanism by which phalloidin acts on the actin filament.  

Pathways to Healthy Arteries

With support of the TeraGrid Pathways program, Shewaferaw Shibeshi explores how to improve diagnosis of deadly arterial disease

Increasing wall-shear stress distribution on the wall of the right coronary artery with increase in hematocrit level (30, 45 & 60). Speaking of PSC consultants Rick Costa and Anirban Jana, who worked with him on this project, Shibeshi says, “Whenever I had a question, they answered it in a day or less.”

Atherosclerosis, better known as hardening of the arteries, is the leading cause of death in the developed world. In the United States alone, more than half of yearly mortality can be traced to arterial disease, with more than 500,000 deaths each year due to heart attacks from arterial blockage.

In response to this major health problem, physicist Shewaferaw Shibeshi has mounted a series of computational fluid dynamics (CFD) studies of blood flow as it relates to arterial disease. He began this work in 2002, as a Ph.D. student at Howard University, and continued it as a professor at Alice Lloyd College in Kentucky. Last year, through the TeraGrid Pathways program, which supports the use of TeraGrid resources by faculty and students at minority-serving institutions, Shibeshi worked with PSC consultants to use Pople, PSC’s shared-memory system, and CFD software called Fluent. With these resources, he simulated blood flow in the right coronary artery to study correlations between the force exerted by flowing blood and “hematocrit” — the proportion of blood volume occupied by red blood cells.

“Blood-flow simulations,”says Shibeshi, “commonly consider the blood's viscosity and density, but other blood components, such as hematocrit, can potentially offer a more accurate and useful explanation of arterial disease process.” Shibeshi’s simulations, which showed a strong correlation between wall-shear stress and hematocrit, confirm findings from clinical and epidemiological studies. His report, submitted this year to Rheology, points out that elevated hematocrit promotes plaque buildup, and that hematocrit level, which can be assessed from blood tests, may be useful for improved diagnosis and treatment of this deadly condition.

Digital Humanities

Through mini-residencies at PSC, scholars are opening new vistas in humanities study

The powerful tools of information technology have begun to open new horizons in many fields that haven't traditionally been involved with computation. In the humanities, the intersection of digital methods with what has historically been a text-based field has spurred the National Endowment for the Humanities (NEH) to support training for humanities scholars exploring digital possibilities.

Through collaboration with NCSA in Illinois and SDSC in San Diego, PSC has provided mini-residencies this year in which scholars from three NEH digital humanities projects experiment with technology and chart long-range goals. Laura McGinnis, PSC manager of education, outreach and training, coordinates this effort. PSC scientists Phil Blood, Thomas Maiden, Nathan Stone and John Urbanic have helped to define objectives and support needs, which include suitable storage media and software development for a “portal” interface and powerful search methods to access information.

Global Middle Ages: A collaboration of the University of Minnesota College of Liberal Arts, School of Music, and Department of History and the Johns Hopkins University Department of the History of Art, the Scholarly Community for Globalization of the Middle Ages (SCGMA) is working to develop a new interdisciplinary community of scholars for study of the Middle Ages. SCGMA aims to create an online infrastructure that will support multiple formats and languages in textual, visual and aural resources. As an example, from a database that would include trade records between two civilizations, they would like to develop a map display that shows how trade routes between these civilizations changed over time.

Humanistic Algorithms: In this project, the University of Southern California's Institute for Multimedia Literary is collaborating with researchers at the University of Illinois, Urbana-Champaign to create a digital archive of multi-media portfolios of faculty and students. The group will use data analytics to extract information from unstructured texts (such as raw textual data like websites) to produce information that can then be used for study to create "meta-analytical" scholarly multimedia.

HistorySpace: A collaboration of international researchers, led by John Bonnett of Brock University and including participants from the United States, Canada, and the United Kingdom, this project aims to incorporate "virtual worlds" as tools to support research and expression in the humanities. Focusing on history and related disciplines, the HistorySpace effort brings historians and computer scientists together to design workflows and tools in mWorlds, an open source virtual world platform.

Optimal Performance

By paying attention to the distance traveled by messages for inter-processor communications, researchers show major gains in computing efficiency

Decomposition of the physical system into chare arrays (only important ones shown for simplicity) in OpenAtom.

Peace, happiness and supercomputing — to perhaps strain for an analogy, there are some things of which you can never have too much. With supercomputing, despite amazing improvements in capability over more than two decades, demand for the resources always exceeds availability. Because of this, computational scientists are constantly pressed for ways — newer, better algorithms — to gain the best possible performance from their research applications. A speedup of 10 percent in software that may require weeks or months of computing to arrive at useful results is enormous.

During the past two years, PSC staff have worked closely with University of Illinois, Urbana-Champaign (UIUC) graduate student Abhinav Bhatele to fine tune and optimize performance on massively parallel systems for OpenAtom, a quantum molecular dynamics application used to study properties of materials and nano-scale molecular structures. OpenAtom implements a highly useful mathematical formulation of quantum mechanics called the Car-Parrinello method, for Italian physicists Roberto Car and Michele Parrinello who introduced it in 1985.

Bhatele developed and tested a way to assign parts of the overall computing job (called chares in the programming model he used, Charm++) to specific BigBen processors. His approach exploits BigBen’s network “torus” topology, also used on other systems, to reduce the number of “hops” that a message between processors has to travel. With OpenAtom, this method yielded speedups of up to 20 percent on 1,024 BigBen processors and up to 50 percent on 2,048 processors. His paper reporting these results won the Distinguished Paper Award at the Euro-Par 2009 conference and also won the outstanding master's thesis award at UIUC.