NCSA NEWS

Where are they now?

Story posted December 12, 2006

Access covers research being conducted by those who use NCSA's computational resources, expertise, and software. But the good work doesn't stop when the presses roll.

Carlos Simmerling, Stony Brook University
Carlos Simmerling and his colleagues caused a stir in 2002, publishing the full folding process of the 20-amino-acid sequence of Tryptophan cage in the Journal of the American Chemical Society. The final structure of the simulated Tryptophan cage was virtually indistinguishable from experimental data, and by 2004 the team was running multiple simulations at the same time to share information and cut down on computing time. (See http://access.ncsa.uiuc.edu/Stories/proteins/)

Now, the team is viewing the moments at which "starter molecules" for HIV are most vulnerable to new drugs. Using NCSA's SGI Altix supercomputer, called Cobalt, they offer insight into the mechanics of HIV protease, a molecule that slices the pre-HIV protein chain into pieces that evolve into a mature virus.

Simmerling's team successfully simulated how HIV protease changes between two forms that already have been determined through experiments. More important, however, is that the group was able to capture the protease in a third, fully open state -- one that had previously been hypothesized but never directly observed. Their work was published earlier this year in the Proceedings of the National Academy of Sciences and the Journal of the American Chemical Society. (See http://comp.chem.sunysb.edu/5_publications.html)

The simulations took 20,000 CPU hours; in real time, the project lasted about three months. Simmerling estimates that it would have taken more than a year to complete the work using his lab's Linux cluster. "It would have been at least six to seven times slower than the Altix, and the cluster doesn't scale as well," he said. "I wouldn't have done it for such a risky project."

Cyrus Madnia, State University of New York at Buffalo
For 15 years, Cyrus Madnia and his team at the State University of New York at Buffalo have used NCSA resources to build direct numerical simulation codes that model complex combustion systems. Simulations like these can be used to show how to alter a flame's characteristics to create more or less heat and less pollution. (See http://access.ncsa.uiuc.edu/Stories/flame/)

Their work currently focuses on nonpremixed flame-wall interaction in which the fluctuation of the temperature and other features of the combustion chamber influence the behavior of the flame and vice-versa. Recently, using NCSA's TeraGrid cluster called Mercury, the team modeled methane combustion using a detailed reaction mechanism with 35 species and 217 elementary reaction steps. Results showed that in the presence of a thermal boundary layer at the wall, the flame chemistry is mainly governed by water reactions that contribute approximately 95 percent of the total heat release rate at the wall.

The team is also exploring ignition dynamics and subsequent flame evolution of hydrogen-enriched methane mixtures. A detailed reaction mechanism and two augmented, reduced mechanisms (11-step and 12-step) were considered. Among other things, the team found that the 11-step model accurately predicts the ignition delay time. At later times, however, the fuel-rich side of the flame predicted by this reduced mechanism exhibits differences from the detailed model.

Results have been published in Combustion Theory and Modelling and Combustion and Flame.

Lubos Mitas, North Carolina State University
Since the 1990s, Lubos Mitas has explored the inner realms of quantum physics--from buckyballs to the molecular electronic structure of silicon nanocrystals. (See http://access.ncsa.uiuc.edu/Stories/electronicstructures/) In April 2006, North Carolina State University's Lubos Mitas and a team from NC State and Arizona State University proposed a new functional form for the efficient description of many-body quantum systems using Quantum Monte Carlo. Wave functions of interacting quantum systems such as electrons in matter are notoriously difficult to calculate despite decades of effort. Quantum Monte Carlo is one of the most productive many-body methods for electronic structure problems.

Their method, published in Physical Review Letters, was developed on NCSA's Tungsten and Copper clusters. The team also computed at NCSA's sister sites, the Pittsburgh Supercomputing Center and the San Diego Supercomputer Center. The method is based on what is called Pfaffian, a mathematical approach to solving skew-symmetric matrices. It allows users to incorporate pair orbitals for both singlet and triplet pairing channels with unpaired one-particle orbitals into a single, compact wave function. With singlet pairing, the correlation between unlike-spin electrons can be described. Triplet pairing does the same for same-spin electron pairs.

Yuriko Renardy, Virginia Tech
Emulsions are significant to a variety of industrial applications, including the development of new materials, wastewater treatment, and pharmaceutical design. Currently, production of these emulsions relies heavily on trial and error. In hopes of improving the state of the art, Yuriko Renardy and her team at Virginia Tech began simulating the nature of drops of liquids in emulsions in 1997. (See http://access.ncsa.uiuc.edu/Stories/dropdeform/)

Previous schemes for determining the distribution of different-sized drops tend to produce errors that do not reconcile themselves no matter how fine-grained the simulation is. Most problematic are false currents within the mixture's flow. The Renardy team's interface reconstruction method, developed on NCSA's Copper supercomputer and previous platforms, eliminates these currents and other errors.

In 2005 and 2006, they used their algorithms to simulate the flow of two viscoelastic liquids. In viscoelastic liquids, drops slide, deform, and change orientation. The team also used the algorithms to predict the evolution of a Newtonian drop in a viscoelastic matrix liquid. Results were published in Rheologica Acta and the Journal of Non-Newtonian Fluid Mechanics.

Benoit Roux, University of Chicago
Previously, Benoit Roux set out to explain how fast ion conduction could take place at a rate near the diffusion limit through the channel and yet maintain a robust ion selectivity. (See http://access.ncsa.uiuc.edu/Stories/ionchannels/) The results of this work were reported in a 2004 issue of Nature. More recently, Roux and his collaborators explored the molecular basis for the activation of these ion channels in response to changes in membrane electrostatic potential, that is, the opening of the portals. This process, referred to as "gating," takes place on a microsecond timescale.

In a 2005 issue of Nature, the team proposed a gating mechanism based on the experimental observation that the transmembrane electric field is actually focused at the charged arginine residues part of the "voltage-sensor" of the channel. It surmises that the arginines do not translocate across the membrane and that the voltage-sensor functions much like a "membrane transporter."

The team uses molecular dynamics simulations on NCSA's Tungsten computing cluster to develop their theories about the function of potassium channels. They also compute at NCSA's sister site, the Pittsburgh Supercomputing Center. They are currently designing suitable computational strategies that will enable them to observe channel gating through these simulations, based on experimentally obtained views of the channel.

"Although the experimentally determined three-dimensional structure of membrane channels yields a wealth of information," Roux says, "the function of these membrane channels is intrinsically dynamical. Theoretical considerations are absolutely necessary for understanding the underlying mechanisms of selective ion conduction and gating."