Entries filed under “Applied HPC”

Alzheimer’s is a progressive and fatal disease that is the number one cause of dementia in the United States, accounting for between 50 and 80% of dementia cases. In 2006 26.6 million people had the disease worldwide, and today it is the 7th leading cause of death in the United States.

There is no cure for Alzheimer’s, and even the causes and mechanisms of disease have yet to be worked out. But scientists are making progress. One of the researchers working on Alzheimer’s is using supercomputers at the Ohio Supercomputing Center to understand what role misfolded proteins may play in the disease process.

In the nucleus of nearly every human cell, long strands of DNA are packed tightly together to form chromosomes, which contain all the instructions a cell needs to function. To deliver these instructions to various other cellular structures, the chromosomes dispatch very small protein fibers – called oligomers – that fold into three- dimensional shapes. Misfolded proteins – called amyloid fibrils – cannot function properly and tend to accumulate into tangles and clumps of waxy plaque, robbing brain cells of their ability to operate and communicate with each other, according to NIH.

…“The exact mechanism of amyloid formation and the origin of its toxicity are not fully understood, primarily due to a lack of sufficient atomic-level structural information from traditional experimental approaches, such as X-ray diffraction, cryoelectron microscopy and solid-state NMR data,” [Jie Zheng, Ph.D.]. explained. “Molecular simulations, in contrast, allow one to study the three-dimensional structure and its kinetic pathway of amyloid oligomers at full atomic resolution.”

Zheng’s research group is developing a multiscale modeling and simulation platform that integrates structural prediction, computational biology and bioinformatics to establish a direct correlation between the formation of oligomers and their biological activity in cell membranes. This research is important for understanding the build-up of protein plaque, how it contributes to the breakdown of cells and how the process might be prevented.

Zheng’s research project recently won an NSF CAREER award, one of the most prestigious the NSF bestows. You can read more about Zheng’s research here.

Engineers at Ohio State University have made progress on research that may enable them to mass produce graphene, a form of carbon that has lots of interesting properties, courtesy of their friendly neighborhood supercomputer

Experts believe that graphene — the sheet-like form of carbon found in graphite pencils — holds the key to smaller, faster electronics. It might also deliver quantum mechanical effects that could enable new kinds of electronics.

…“Graphene has huge potential — it’s been dubbed ‘the new silicon,’” said Padture, who is also director of Ohio State’s Center for Emergent Materials. “But there hasn’t been a good process for high-throughput manufacturing it into chips. The industry has several decades of chip-making technology that we can tap into, if only we could create millions of these graphene structures in precise patterns on predetermined locations, repeatedly. This result is a proof-of-concept that we should be able to do just that.”

Daydreams of potential uses include chemical sensors, and in future computer chips

Researchers have shown that a single sheet, or even a few sheets, of graphene can exhibit special properties. One such property is very high mobility, in which electrons can pass through it very quickly — a good characteristic for fast electronics. Another is magnetism: magnetic fields could be used to control the spin of graphene electrons, which would enable spin-based electronics, also called spintronics.

What’s the computing angle?

In computer simulations, they found that each material [tested as a potential substrate] interacts differently with the graphene. So success might rely on finding just the right combination of substrate materials to coax the graphene to break off in one or two layers. This would also tailor the properties of the graphene.

Those simulations were run at the Ohio Supercomputer Center using the Vienna Ab-initio Simulation Package. From OSC’s web site

“The calculations are computationally very demanding for the systems under consideration due to their size and complexity, and they couldn’t have been done without our allotment at the Ohio Supercomputer Center,” Windl explained. “Based on our initial success with these computer simulations, we currently model adhesion on different substrates along with the resulting electrical transport through the graphene to optimize the stamping process and the resulting devices.”

More in the article at the link above, along with a pointer to a detailed paper in the journal Advanced Materials.

OSC sent us news this week about a researcher at The Ohio State University who has been using OSC’s supers to develop a new application to track and analyze the genetic codes of swine flu virus as it travels the world, moving from host to host and mutating to resist antiviral drugs.

In a journal article published in the April 2010 online issue of Cladistics, Daniel Janies, Ph.D., explains how Supramap (http://supramap.osu.edu) was created to track the avian influenza virus (H5N1) and, more recently, to monitor the H1N1 virus. Cladistics refers to the scientific classification of living organisms, based on common ancestry, into evolutionary trees. Evolutionary trees are used by many researchers studying infectious diseases to understand the geographic and host origins of pathogens and how the pathogens change over time. Supramap puts phylogenies in a geographic context as well.

“The integration of our core phylogenetic reconstruction codes with Supramap has allowed an entirely new way to view linked evolutionary and geographic information,” said Ward Wheeler, a coauthor of the article and curator-in-charge of scientific computing at the American Museum of Natural History (AMNH). “The Supramap tool set has broad utility not only in tracking human disease in time and space, but historical patters of biodiversity and global biotic changes.”

This is one of those applications that just screams for someone to think hard about how to make it available to non-HPC specialists.

Janies and his colleagues used a small cluster computer at OSU to beta-test the Supramap application, which has been developed through a grant from the Defense Advanced Research Projects Agency (DARPA).

The research team also adapted the Supramap code to function smoothly on the OSC’s flagship IBM Cluster 1350 “Glenn” system, which features 9,500 cores and 24 terabytes of memory. They now are working with the Center’s staff to finish development of a Web interface to provide easy Internet access to the application by scientists and public health officials.

Two bio articles related to HPC are making their way around the interwebs this week. The first is about using supercomputers to identify molecular structures with the potential to have value as new medicines. From Physorg.com

Previous methods to identify these molecules have emphasized searching for fragments that can attach to one hot spot at a time. Finding structures that attach to all of the required hot spots is tedious, time-consuming and error-prone.

Ohio State University researchers, however, have used computer simulations to identify molecular fragments that attach simultaneously to multiple hot spots on proteins. The technique is a new way to tackle the fragment-based design strategy.

“We use the massive computing power available to us to find only the good fragments and link them together,” said Chenglong Li, assistant professor of medicinal chemistry and pharmacognosy at Ohio State and senior author of a study detailing this work.

…”My method reconstructed what pharmaceutical companies have already done,” he said. “In the future, we’ll apply this technique to protein targets for diseases that remain challenging to treat with currently available therapies.”

Then, ScienceBlog.com reports that researchers at the Virginia Bioinformatics Institute (VBI) and the Department of Computer Science at Virginia Tech are using HPC to find small genes that have been missed by scientists as they identify DNA sequences

Using an ephemeral supercomputer made up of computers from across the world, the mpiBLAST computational tool used by the researchers took only 12 hours instead of the 90 years it would have required if the work were performed on a standard personal computer.

The new study, reported in the journal BMC Bioinformatics, is the first large-scale attempt to identify undetected genes of microbes in the burgeoning GenBank DNA sequence repository that contains over 100 billion bases of DNA sequence. The genes uncovered may have important functions in the cell, but those functions need to be established by further experiment.

…João Setubal, associate professor at the Virginia Bioinformatics Institute and the Department of Computer Science at Virginia Tech, commented: “Scientists have known for a long time that publicly available databases of genomes have inconsistencies, errors, and gaps. Some genes are labeled with the wrong function and for others the function is unknown. But nobody had done a systematic study to verify how many genes were simply undetected. This is what we did in our study — discover the number of microbial genes that are under the radar.”

You can read the full paper describing that work here.

HPC plus collider equals new piece of reality

Late last week Berkeley Lab announced that researchers using NERSC’s HPC systems and the Brookhaven National Laboratory’s Relativistic Heavy Ion Collider (RHIC) have detected and confirmed the first-ever antimatter hypernucleus.

Translated, the newly detected “antihypertriton” means a nucleus of antihydrogen containing one antiproton and one antineutron—plus one heavy relative of the antineutron, an antilambda hyperon.

What was the role of the 100,000 hours of computation in all this? Set your WABAC Machine for the Big Bang at the beginning of…well…everything

This very hot cosmic stew of free floating fundamental particles, including quarks, antiquarks and gluons is known as the quark-gluon plasma. As the universe expanded and cooled, the quarks recombined in a variety of ways to make protons and neutrons (consisting solely of up and down quarks), hyperons (which contain strange quarks) and all of the associated antiparticles. Because quarks and antiquarks exist in equal numbers in the quark-gluon plasma, the cooling gas produces both matter and antimatter. Eventually, a small fraction of these particles combined to form light nuclei and their antiparticles like the antihypertritons detected by the STAR collaboration. To identify this hypernucleus, physicists used supercomputers at NERSC and other research centers to painstakingly sift through the debris of some 100 million collisions.

The team also used NERSC’s PDSF system to simulate detector response. …”These simulations were vital to helping us optimize search conditions such as topology of the decay configuration,” says Zhangbu Xu, a physicist at Brookhaven who is part of the STAR collaboration. “By embedding imaginary antimatters in a real collision and optimizing the simulations for the best selection conditions, we were able to find a majority of those embedded particles.”

More in the story.

Got an email this week about a new @Home project, this one aimed at advancing nuclear fusion research

Producing energy in a sustainable, safe and  environmental friendly way is one of the main challenges of our time. Nuclear Fusion is one of the promising energy sources, but  the technology is difficult to master.  It could provide endless energy – estimated to be sufficient  for  about 2 billion years, there is no CO2  produced and no nuclear waste. Fusion reactors are also inherently safe, because the reaction automatically stops when something goes wrong. Currently Europe, and other countries from all over the world are working together to develop and  build  nuclear fusion reactors. To predict and optimise the behaviour of nuclear fusion reactions in these  machines, a lot of computer calculations are needed.  Everyone can now help with solving this problem by donating unused computing time to the  EDGeS@Home project. The otherwise wasted computing time can then be used by scientists in Europe to run software called ISDEP  designed to do part of these reactor simulations. ISDEP has been ported by the EDGeS project to the EDGeS infrastructure and is used in EDGeS@Home. In this straightforward way, this ‘green’ approach of information technology and computation make the Fusion research itself environmental friendly now.

…By donating unused computing time to ISDEP, citizens can help solving the world’s energy problem. The only thing they have to do is to download a small  easy-to-install  Desktop Grid programme, that will take care unused computing time is used by ISDEP. Contributing to a Desktop Grid such as EDGeS@Home is secure and safe: the Grid software itself is inherently safe and secure. The scientific programmes have gone through an extensive validation process to be absolutely sure they cannot harm a citizen’s computer in any way.

The project officially launches on Mar 30, but I think you can actually download the client now and start working. More at the web site.

Those insideHPC readers who are motorsports fans, especially fans of [what I consider to be] the pinnacle of all motorsports, Formula One racing; are probably familiar with the time and expense race teams put into designing and tweaking their respective platforms.  Traditionally, much of the black art of chassis dynamics was performed through lengthy sessions in expensive wind tunnels.  Chassis manufacturers and race teams simply trusted the empirical results elicited through blasting air at high speeds over their multi-million dollar carbon toys and utilizing computational methods as broad guides to reasonable platform mock ups.  However, this might be changing.

According to and article posted on Planet-F1, the Virgin Racing Formula One team will make history this season, before even beginning a single race.  Their cars were designed solely using computational methods and zero wind tunnel testing.  Nick Wirth, Virgin Racing Technical Director, is the leading the effort.  Previously, he directed the technical aspects of a CFD-centric design for team Acura in the American Le Mans series.  These efforts paid off as Acura took first in the LMP1 and LMP2 classes last year.

The point people seem to be missing is the fact CFD is just a tool, like a scale model or whatever, and it gives you a set of answers to a set of questions that you put into it,” said Wirth.

All I want to ensure is wherever we are in Bahrain (opening grand prix on March 14) we then show a rate of development during the year with this process which is faster.

“In Bahrain, you cannot say whether the process has worked or not. It will be about the rate of development, which is the most important thing in the world to a racing team.

The move by the Virgin Racing team certainly has its skeptics within the realms of Formula One design. Adrian Newey directs technical operations for the Red Bull racing and Dr. Mike Gascoyne, who is now with Lotus Racing, are both skeptical of completely dropping empirical testing.

I think CFD is a very powerful tool, there is no doubt about it, and it is another way of simulating the real environment, but it still has pitfalls,” remarked Newey.There are some areas that CFD physically doesn’t capture as well as a wind tunnel, like basic aerodynamic properties.

You look at BMW when Albert II was announced as one of the world’s biggest supercomputers dedicated just to their CFD,” said Gascoyne.  If you look at Renault, they built their environmentally-friendly CFD centre, with a huge computing resource.  I don’t think these guys are idiots, and they also have windtunnels.

Motorsports may sound somewhat benign as a place to innovate in the realm of computational fluid dynamics.  However, remember that these teams thrive on the millions of dollars in sponsorships anchored by a winning race team.  In a sport governed by hundredths of seconds in elapsed time, the slightest engineering edge could potentially propel a team onto the podium.

If you’re interested in reading more, check out the article here.