Blazingly fast IBM-designed supercomputer is helping Los Alamos researchers focus on promising HIV vaccine targets
LOS ALAMOS, N.M.—The biggest worry on the mind of the humanimmunodeficiency virus—well, if it had a mind, anyway—these days might not bethe latest antiretroviral therapies, but rather a member of the periodic tableof the elements going by the name of silicon. And that's because acutting-edge, petascale supercomputer called Roadrunner, developed by IBM inpartnership with the Los Alamos National Laboratory and the National NuclearSecurity Administration, is being put to work analyzing vast quantities ofgenetic sequences from HIV-infected people in the hope of zeroing in on possiblevaccine target areas.
The vaccine research is part of Los Alamos NationalLaboratory's role in the international Center for HIV/AIDS Vaccine Immunology(CHAVI) consortium, and involves studying the evolutionary history of more than10,000 sequences from more than 400 HIV-infected individuals. To do this, twoLos Alamos researchers—physicist Tanmoy Bhattacharya and HIV researcher BetteKorber—are using samples taken by CHAVI across the globe, from both chronic andacute HIV patients, to create an evolutionary genetic family tree, known as aphylogenetic tree. This will allow researchers to look for similarities in theacute versus chronic sequences that may identify areas where vaccines would bemost effective.
According to Korber, once they can identify common featuresof the transmitted virus, that will be a breakthrough in attempts to create avaccine that enables recognition the original transmitted virus before thebody's immune response causes the virus to react and mutate.
But none of this would be possible, they note, without theprocessing power of the IBM-designed Roadrunner at Los Alamos, which on May 26,2008, exceeded a sustained speed of 1 quadrillion (1015) calculations persecond, breaking the so-called petaflop barrier. Shortly after thatachievement, Roadrunner was named the world's fastest supercomputer by theTOP500 organization at the June 2008 International Supercomputing Conference inDresden, Germany.
"The petascale supercomputer gives us the capacity to lookfor similarities across whole populations of acute patients," Bhattacharyaexplains. "At this scale, we can begin to figure out the relationships betweenchronic and acute infections using statistics to determine the interconnectingbranches—and it is these interconnections where a specially-designed vaccinemight be most effective."
"DNA sequencing technology [is] currently beingrevolutionized, and we are at the cusp of being able to obtain more than100,000 viral sequences from a single person," adds Korber. "For this new kinddata to be useful, computational advances will have to keep pace with theexperimental, and the current study begins to move us into this new era."
That IBM is playing a key role in this work is not assurprising at it might first seem, given the computing giant's increasinglyactive role in life sciences research. As reported in the November 2009 issueof ddn, IBM is making waves with itsconcept of the DNA Transistor, a device very much like a typical siliconmicroprocessor-style chip, with a 3-nanometer-diameter hole through which a DNAmolecule would be passed in order to read it step-by-step quickly, possiblyoffering a means to achieve the elusive genetic sequencing goal that everyonecalls the "$1,000 genome."
IBM is involved with a variety of other life sciencesprojects as well—both internally and in partnerships and collaborations withother organizations—such as bioinformatics and pattern discovery in the area ofmolecular biology, nanoscale dewetting transition in protein folding andresearch into quantum computing.
"Some years ago, we said that science was making thetransition from the century of physics to the century of life sciences," saysDave Turek, vice president of supercomputing at IBM. "Because of thatrealization, IBM formed a computational biology center at its T.J. WatsonResearch Center that, among other things, uses supercomputing, the developmentof algorithms and the use of in silicomodeling to solve a plethora of issues, including challenges in drug discoveryand development."
Using in silico toolsin pharmaceutical research offers obvious benefits in terms of doing moreaccurate modeling of efficacy and safety before actually having to spend moneyon animal or human studies, as well as the ability to potentially model large populationswithout having to use actual people. But more than that, computers canpotentially model pathogens, biological processes and therapeutic agents attime scales that are very tiny and virtually impossible for humans to deal withotherwise.
"We humans operate at clock time—what we see on that wallclock or our watches," Turek says. "But biological processes can take place attime scales much smaller than that—atomic scales, and you cannot see that in atypical lab setting."
To get down to that level of detail and truly examineimportant issue like protein folding, he says—or even to adequately modelpopulations or pathogen mutations over long spans of time such as decades orcenturies—requires a petascale supercomputer.
"The implications of this, now that we have that kind ofcomputing capability, are immense," Turek notes, particularly with regard tosomething like vaccine development, whether for HIV or for influenza. "In the1918 influenza pandemic, it was lethal, but it tended to stay concentrated incommunities. In the modern age, with people able to travel so far and so fast,we don't have the luxury of a virus staying contained for a long time while wefigure out after the fact what kind of vaccine we need. A lethal enough flucould kill millions by the time we had a vaccine.
"Modeling likely mutations over a long period of time, usinga computer-like Roadrunner," he continues, "could enable us to predict likelystrains and their evolutionary steps, and be ready with enough knowledge todevelop vaccines as the viruses emerge, or perhaps even before they do."
On the HIV/AIDS front at Los Alamos, with Roadrunner to aidthem, Korber and her team currently are designing three vaccines to target HIV.Their vaccine model is based on a mixture of synthetic proteins that addressthe virus' evasive nature—its ability to avoid triggering an immune response;its ability to mutate quickly, thereby increasing its drug resistance; itsdefensive cloak of sugar molecules that prevents antibodies from blocking the HIVproteins used to invade the cell; and more. Animal tests are underway,reportedly with promising results thus far and human trials expected to beginsoon.
Greater than the sum
By Jeffrey Bouley
ARMONK, N.Y.—What makes the Roadrunner petascalesupercomputer at Los Alamos National Laboratory in New Mexico isn't just thatit can do such an extraordinary number of calculations so quickly. It's alsothe fact this computer was built in a very novel manner.
IBM didn't simply aim to build a bigger supercomputer, notesDave Turek, vice president of supercomputing at IBM. Instead, it built a hybridsystem, which gets its processing power from 12,240 IBM PowerXCell 8i CellBroadband Engine processors that are derived from the same microchips thatpower today's most popular videogame consoles.
In addition, 6,562 AMD Opteron Dual-Core processors performbasic computer functions, freeing the IBM PowerXCell 8i chips for themath-intensive calculations that are their specialty.
This hybrid architecture, which optimizes the strength ofmultiple types of processors, is analogous to that of a hybrid car, andoffering similar benefits in terms of not only efficiency, but also energyusage. If Roadrunner had been built with standard x86 chips alone, the systemreportedly would have been significantly larger and would have required muchmore power.
"The power problem is one that affects our industryubiquitously," Turek notes. "Hybridization allows us to pick and choose andmanage the power profile of the system to get an extremely high level ofperformance per watt. If we had tried to build Roadrunner without a hybridapproach, we would have blown both the space allocation and the power budget atLos Alamos by far."
Roadrunner races to its next duty
By Jeffrey Bouley
LOS ALAMOS, N.M.—The work on HIV vaccine research isn't theonly thing the Roadrunner supercomputer has been up to. During the samesix-month period that it was doing that, which ended in September, Los Alamosresearchers had Roadrunner, working on 10 unclassified projects. Some of theother projects besides HIV include understanding the nature of dark matter anddark energy, understanding the non-linear physics of high-power lasers,examining how shockwaves cause materials to fail, and unraveling how and whysome stars explode.
But in the end, while Roadrunner has advanced science in anumber of areas, all of these projects were run on the supercomputer as a kindof "shakedown" process so thatscientists could optimize the way large codes are able to run on the machineand ensure that it could handle huge workloads.
Now Roadrunner is in the process of transition to classifiedcomputing for its intended mission: to assure the safety, security, andreliability of the U.S. nuclear deterrent.