Three miles from the main campus, Princeton’s high-performance computers hum undisturbed, cranking out projections of what happens when a neutron star encounters a black hole — things don’t go well for the neutron star — working out how trees know when it is safe to put out their spring leaves, and designing drug candidates for treating inflammatory diseases.
The assortment of projects these machines tackle is a testament to the role of “big data” in scientific research as well as to Princeton’s pioneering philosophy of making high-performance computers available to any University researcher in need of these powerful resources.
“Princeton’s approach is really unique,” said Jeroen Tromp, the Blair Professor of Geology and director of the Princeton Institute for Computational Science and Engineering (PICSciE), which oversees the new facility in conjunction with the University’s Office of Information Technology. “At most universities, researchers work department-by-department or individually to get the computing resources they need.”
In November 2011, Princeton consolidated its silicon workforce — consisting of five high-performance clusters located in various locations around campus — in a state-of-the-art facility on Princeton’s Forrestal Campus. The facility is the centerpiece of Princeton’s innovative plan to provide robust computing resources to all faculty members and researchers.
Modeling the creation of gravitational waves
They start as tsunamis out in space but by the time they reach Earth, gravitational waves are mere ripples. Predicted by Einstein, these waves have never been detected, but three observatories — in the United States, Japan and Germany — are actively seeking proof they exist.
Physics professor Frans Pretorius uses the University’s high-performance computers to model the emergence of gravitational waves from their origins in the violent collisions of black holes and other extremely dense objects. With assistance from graduate student William East and former postdoctoral researcher Branson Stephens, now at the University of Wisconsin-Milwaukee, Pretorius is using the computers to run simulations of these collisions and follow the resulting waves through time to see what they will look like by the time they reach Earth.
Modeling these waves as they reach Earth could inform gravitational wavehunters what signatures to look for in otherwise noisy data, as the observatories are sensitive to a variety of other disturbances, including natural seismic activity. “If we know what the waves coming from these collisions should look like, then we will know what to look for in the data collected at these observatories,” Pretorius said.
Pretorius’ simulations reveal what can happen when a neutron star encounters a black hole. At first the neutron star and the black hole circle each other in a slow elliptical dance. As the neutron star succumbs to the pull of the black hole’s gravity, the dance speeds up and the two get closer and closer until the star falls into the black hole and is ripped apart.
The result is a tremendous release of energy, Pretorius said. “It is like taking a tenth of the sun and converting it entirely into gravitational wave energy within a few milliseconds.” The work, which was funded by the National Science Foundation (NSF) and the Alfred P. Sloan Foundation, was published in Astrophysical Journal Letters in July 2011.
Modeling this process is computationally intensive. “We cannot just run one simulation to understand what is going on,” Pretorius said. “For example, we need to vary the spin of the black hole, the mass ratio, the equation of state of the neutron star — there are a huge number of parameters that we need to explore.”
Pretorius’ models of what this energy looks like by the time it reaches Earth will help gravitational-wave observatories in their search. And he said there is another perk of detecting gravitational waves — confirming the existence of black holes. “So far we have inferred that black holes exist,” Pretorius said, “but detecting gravitational waves would be the first real observation of them.”
Tracing trees’ carbon dioxide levels
If black hole collisions seem out of this world, then the research of David Medvigy is something most of us can relate to — the climate. Medvigy, an assistant professor of geosciences, uses high-performance computers to model the Earth’s climate, and he has noted that many of these models are missing something big: trees.
Because modeling the Earth’s climate is so complex, trees are often represented as a single species or as a single, gigantic leaf covering the landscape. Yet trees are significant carbon storage mechanisms, so depicting them accurately is important, Medvigy said. To address this issue, he set out to explore the effect of deciduous trees, which drop their leaves every winter and regrow them each spring, on carbon dioxide levels in the atmosphere.
“We all know that trees put out their leaves in springtime, and this new growth results in the uptake of carbon dioxide,” Medvigy said. “My objective was to write down an equation that would describe leaf extension.”
The timing of leaf extension, or spring budburst, determines when trees begin taking up carbon dioxide, which in turn influences how much carbon dioxide is taken up by trees each year.
Most models used temperature as the driver of budburst — when the weather turns warmer, the trees respond by putting out new leaves. But the models were not very good at predicting when this would occur, so Medvigy reasoned that there was another factor besides temperature that was helping trees “decide” that winter was over.
By trying a number of parameters in their computerized model, Medvigy and his colleagues found that, in addition to temperature, trees also take into account how many cold days have passed. By taking the cumulative number of cold days as a rough estimate of how far winter has progressed, trees may be likely to avoid exposing new, fragile leaves to frost by waiting until a certain number of cold days have elapsed, the team found.
The Princeton team included Su-Jong Jeong, a postdoctoral researcher in the Program in Atmospheric and Oceanic Sciences, a collaboration between Princeton and the Geophysical Fluid Dynamics Laboratory (GFDL), a climate studies lab located on Princeton’s Forrestal Campus and administered by the National Oceanic and Atmospheric Administration (NOAA).
Medvigy and Jeong added their budburst simulation to an existing climate model created by Elena Shevliakova and Sergey Malyshev, both affiliated with Princeton’s Department of Ecology and Evolutionary Biology and GFDL. The researchers found that their model showed carbon uptake from deciduous trees happening about 11 days earlier than previous models — representing a 6 percent increase in the amount of carbon they took in during the year — and that the model compared well to actual tree behavior at forest sites across the United States. The work was funded by NOAA and published in the Journal of Geophysical Research in March 2012.
Designing better drug candidates
Princeton’s high-performance computers also provide new knowledge at the microscopic level of the protein. These complex molecules do most of the jobs in the body, including providing structure to cells, giving drugs access to cells and forming antibodies that detect and eliminate disease-causing microbes.
One thing all proteins have in common is that they are composed of bulky chains of amino acids. These amino acids fold than the most elegant origami.
Understanding how proteins fold could enable researchers to devise new treatments for a variety of diseases. Developing models of folding and designing new proteins is the research focus of Christodoulos Floudas, the Stephen C. Macaleer ’63 Professor in Engineering and Applied Science at Princeton.
Floudas’ early models were some of the first to depict amino acid chains as being flexible rather than having fixed backbones. Using these models, it is possible to predict, starting from only the amino acid sequence, the structure of the folded 3-D protein. The computations can take a month of computer time.
Floudas and his team are modeling protein structures that are known to be important in governing certain immune system responses. Their work has the potential to lead to new routes of control- ling inflammation in diseases such as asthma, rheumatoid arthritis, stroke and sepsis.
The team, which included collaborators Dimitrios Morikis and his team at the University of California-Riverside, and Princeton chemical and biological engineering postdoctoral researcher Christopher Kieslich, used molecular-dynamics simulations — computer programs that mimic the physical movements of atoms and molecules — to create a 3-D model of a protein structure called the C3a receptor, which sits on the surface of human cells and plays a critical role in regulating immune responses. Then former Princeton graduate students Meghan Bellows-Peterson, Class of 2011, and Ho Ki Fung, Class of 2008, designed short portions of proteins called peptides that bind to the C3a receptor and either block or enhance aspects of its activity.
After synthesizing the predicted peptides, collaborators led by Peter Monk at the University of Sheffield Medical School in England and Trent Woodruff at the University of Queensland in Australia tested the peptides and showed that they behaved as predicted in cells. The team published their results in the April 2012 issue of the Journal of Medicinal Chemistry.
“This is an example of how computer modeling and optimization can lead to novel therapies that have the potential to make a difference in patient treatment,” Floudas said. Funding for Floudas’ work came from the National Institutes of Health, the NSF and the U.S. Environmental Protection Agency.
Bellows-Peterson, Meghan L., Ho Ki Fung, Christodoulos A. Floudas, Chris A. Kieslich, Li Zhang, Dimitrios Morikis, Kathryn J. Wareham, Peter N. Monk, Owen A. Hawksworth and Trent M. Woodruff. 2012. “De Novo Peptide Design with C3a Receptor Agonist and Antagonist Activities: Theoretical Predictions and Experimental Validation.” J. Med. Chem., Vol. 55, no. 9: 4159-68.
Jeong, Su-Jong, David Medvigy, Elena Shevliakova and Sergei Malyshev. 2012. “Uncertainties in Terrestrial Carbon Budgets Related to Spring Phenology.” J. Geophys. Res., Vol. 117, G01030.
Stephens, Branson C., William E. East and Frans Pretorius. 2011. “Eccentric Black Hole- Neutron Star Mergers.” Astrophys. J., Lett. Vol., 737: L5.