Illinois Simulation Software Contributes to Breakthrough in Understanding HIV Capsid
7/16/2013 10:19:00 AM
Recent research on the structure of the HIV capsid was accomplished using simulation software developed at the University of Illinois. A decades-long collaboration involving CS Professor Laxmikant “Sanjay” Kale and Physics Professor Klaus Schulten has centered on developing state-of-the-art software tools for biomolecular modeling. NAMD (NAnoscale Molecular Dynamics) was first developed by Kale, Schulten, and co-workers nearly 17 years ago.
Using NAMD, Schulten and his colleagues determined and illustrated the structure of the HIV capsid—research that was featured on the cover of the journal Nature.
Researchers throughout the world use NAMD to simulate biological machinery at the atomic level on parallel supercomputers and clusters. This scalable software enables simulations of very small molecular systems of the size from a few hundred thousand atoms to larger ones of perhaps a hundred million atoms. “At that level, you can actually see how various biological phenomena happen,” said Kale.
NAMD is written using Charm++, a parallel programming system developed by Kale’s Parallel Programming Laboratory. Charm++ is likewise nearly 20 years old and has evolved in conjunction with NAMD. “We leverage the features of Charm++ to make NAMD an effective, scalable program,” said Kale. “At the same time, NAMD influences the design of newer features in Charm++. So they kind of co-evolved.” Jim Phillips, from Schulten’s research group, is the lead developer of NAMD. The NAMD development team includes many students and staff from Kale’s group.
Over the years, NAMD has become very popular with researchers. Kale and his research group regularly update NAMD and Charm++ with new capabilities and to ensure performance on new machines and systems. Vendors of parallel machines also consult with Kale’s research group. “When parallel machine vendors sell a new machine, their customers want to be sure that NAMD works well on the machines,” said Kale. “So those companies come to us for help, whether it be IBM or Cray or NVIDIA.”
As scientific modeling has progressed, a need has grown for larger and larger computers to enable the simulations needed by researchers. When the National Science Foundation (NSF) first called for proposals for a petascale computing system (a call that led to the current Blue Waters facility on the U of I campus), one of the conditions for a successful proposal was to demonstrate scalable performance using NAMD.
When Blue Waters was ready to run NAMD, there were several challenges that Kale and his team had to address (as shown in a supercomputing 2012 paper) . Performance bottlenecks were discovered, and some memory issues had to be overcome. “This happens all the time,” said Kale. “For every new machine, for every new scale, you have to overcome new challenges. It can be like a game of Whac-A-Mole.”
Breakthroughs such as the recent HIV capsid results make the years of effort worthwhile for Kale and his team. “This shows how computer science research, driven by applications needs, can develop something that can have an impact,” he said.
In 2002, NAMD received the Gordon Bell Award. “The Gordon Bell Award is one of the prestigious awards in parallel computing,” said Kale. “They are given each year to a program that demonstrates maximum performance in some fashion.” In 2012, Kale and Schulten received the IEEE Computer Society Sidney Fernbach Award “for outstanding contributions to the development of widely used parallel software for large biomolecular systems simulation.”
Kale is co-PI of the Theoretical and Computational Biophysics group in the Beckman Institute. This group, an NIH Center for Macromolecular Modeling and Bioinformatics, is led by Schulten and has been in existence for more than 20 years.