The National Resource for Biomedical Supercomputing at the Pittsburgh Supercomputing Center and D.E. Shaw Research have agreed to extend a National Institutes of Health-funded program that grants researchers free access to a supercomputer called Anton, designed for molecular-dynamics simulations of biomolecular systems, beyond its scheduled end date.
Last September, the National Institute of General Medical Sciences awarded a $2.7 million “Grand Opportunities” stimulus grant to PSC to host Anton, which DESRES developed with the aim of making it freely available for non-commercial research use by universities and non-profit institutions (BI 05/05/2010).
However, the program, originally expected to end on Aug. 31, has been extended for nine more months following a successful first round.
To that end, in May the NRBSC issued a new request for proposals for a second phase urging both current and new research projects to apply for time on the machine.
The deadline for submissions was June 23.
Markus Dittrich, a biomedical scientist at the NRBSC who manages Anton's operations at PSC, told BioInform that the second round of projects will be supported with funds leftover from the original NIGMS grant.
When the project began, "we weren’t really sure how large the cost would be to us to run the machine," he explained.
However, since the machine "has been running extremely well" and at a lower cost than estimated previously "we decided to use the remaining funds to make it available and really have the scientific community benefit as much as possible."
In the first round of awards, 47 research groups were selected to use the machine for their research project.
For the second phase, Dittrich said the NRBSC received more than 80 proposals. However, it is looking to support about the same number of projects as the previous round.
He also said the National Research Council of the National Academies of Science convened a panel that met last week to select the projects that will have access to Anton in this phase, which will kick off at the end of the year.
"We will do about 10 to 15 awards that are [around] 100,000 node hours ... and maybe 25 to 30 awards that are around 50,000 node hours," he said.
According to the RFP, the final allocation decisions should be announced in late August.
DESRES representatives declined BioInform's request for comment.
Bigger Simulations in Microseconds
In one project, researchers from the University of Illinois, Urbana-Champaign, used Anton to simulate the folding process of a protein called lambda6-86 "at full atomistic detail" with "all protein atoms and water molecules included," Martin Gruebele, a professor of biophysics and computational biology at UIUC and one of the investigators on the project, explained in a statement.
With 80 amino acids, this protein is more than twice as large as the largest proteins whose folding activities have been simulated and published previously, the researchers said.
Gruebele added that the simulations "revealed interesting dynamics in one of the alpha helices that weren’t observed by experiments before" and suggested possible experiments that will help the team probe deeper into the folding mechanism of this protein.
Klaus Schulten, a professor of biophysics also at UIUC, is currently using Anton to simulate the process by which nascent proteins are threaded from the ribosome into the cellular membrane.
His team has applied for additional time on Anton in the second round of allocations.
In another project, Emad Tajkhorshid, an associate professor of computational biology and biophysics also at UIUC, used Anton to simulate structural changes in membrane transporter proteins — molecular passageways that open and close to move biomolecules such as neurotransmitters in and out of cells.
“Before Anton ... we could simulate maybe 100 nanoseconds of protein motion," he said in a statement. However, on the supercomputer, the team was able to "run several microseconds of simulation — more than 100 times longer in biological time.”
Improving Infrastructure for Molecular Dynamics
Last October, DESRES researchers published a paper in Science that shed some light on Anton's simulation capabilities.
In the article, the team notes that while molecular simulations are useful tools for understanding protein-folding activities and conformation changes in their folded states, their utility has been limited by computational constraints.
That’s because "many biological processes involve conformational changes that take place on time scales between 10 microseconds and 1 millisecond" while current supercomputers limit simulations to about "1 microsecond of simulated biological time" and take a lot longer to run.
For instance, they wrote, the longest previously published all-atom simulation of a protein was 10 microseconds and required more than 3 months of supercomputer time to complete.
Anton was specially designed to speed up "the execution of such simulations, producing continuous trajectories as much as 1 ms in length" and enable research efforts in "protein folding and the interconversion among distinct structural states of a folded protein," the researchers wrote.
Anton differs from other systems used for molecular dynamics simulations — such as MDGRAPE-3, developed by Japan's Institute of Chemical Research, RIKEN — because it does its computations solely on specialized application-specific integrated circuits rather than dividing them between these ASICs and general purpose host processors.
Using Anton, the investigators were able to "perform continuous, all-atom molecular dynamics simulations of proteins in an explicitly represented solvent environment over periods as much as 100 times longer than was previously feasible," they wrote.
In one example described in the paper, the team ran two100-microsecond simulations of a protein dubbed FiP35 as it went through 15 folding and unfolding transitions.
Additionally, they were able to observe the sequence of events between the folded and unfolded protein as well as explore possible explanations for its folding route.
Another recent effort focused instead on improving protein-simulation software was developed by researchers at the Massachusetts Institute of Technology along with collaborators at Boston College and McGill University in Montreal, Quebec.
Based on ensemble modeling, the approach first samples the complete conformational landscape of large proteins based on sequence data alone, and then builds a "coarse-grain" representation of the protein's energy landscape that is used to model intermediate folding states and, subsequently, the folding process.
The researchers claim that their method can enable protein-folding simulations for very large proteins on a single central processing unit in minutes — a process that otherwise requires hundreds of thousands of CPU-hours with all-atom molecular dynamics simulations (BI 03/25/2011).
Have topics you'd like to see covered in BioInform? Contact the editor at uthomas [at] genomeweb [.] com