How MSU is working to future-proof chemical discovery
Michigan State University researchers are leading a new grant project funded by the National Science Foundation at more than $3.8 million to keep computational chemistry’s infrastructure humming.
That probably sounds a little intimidating, but computational chemistry is kind of like an internet service provider. They both impact modern life but aren’t typically at the front of most people’s minds — that is, until they’re not as fast or reliable as we expect them to be.
When there are hiccups or inefficiencies in the computer systems that simulate chemistry, however, it probably doesn’t directly affect as many people as, say, the internet going down. Still, the folks working with computational chemistry have very important jobs. They’re designing chemical products and processes to build a more affordable, more sustainable and healthier future.
“For example, basically every pharmaceutical company in the world has a computational chemistry group,” said Kenneth Merz, Joseph Zichis Endowed Chair and a University Distinguished Professor with joint appointments in the Department of Chemistry and the Department of Biochemistry and Molecular Biology in the MSU College of Natural Science.
With support from NSF’s Office of Cyberinfrastructure and Division of Chemistry, Merz and a multidisciplinary team are working to modernize and future-proof tools and techniques in computational chemistry that will help accelerate discoveries in many diverse fields.
“What people want to be able to do is design really useful things from first principles,” Merz said. “This could be for improving batteries or designing enzymes to break down plastics that don’t naturally biodegrade. It could be designing a new drug from scratch.”
Joining Merz on the leadership team are H. Metin Aktulga from the MSU College of Engineering, Darrin York of Rutgers University and Andreas Goetz of the San Diego Supercomputer Center at the University of California, San Diego.
A future preserved in Amber
Together, the researchers are working with a suite of simulation software packages called Amber. Amber has been around since the 1970s and, thanks to scientists including Merz and York, it has become a workhorse of computational chemistry.
An illustration of the fields of chemistry that use Amber, represented at the center of the image as graphics processing unit encased in amber. Credit: Associate Research Professor Şölen Ekesan/The York Lab/Rutgers University
With the new NSF grant, the team is working to make Amber more powerful using what are known as free energy simulations. The team is also coupling those improvements with other modern advances in both software and hardware. That means making Amber seamlessly compatible with today’s machine-learning algorithms and fastest supercomputers — as well as the advances tomorrow might bring.
Again, the technical complexity can be intimidating, but the goal is straightforward. Going back to the example of an internet provider, it’s like ensuring fast, reliable Wi-Fi no matter what type of device people are using or which version of a browser they have.
“We’re going to deliver state-of-the-art software that you can use on big machines that’s flexible and easy to use,” Merz said. “This will allow anyone to come in and do their job with good accuracy.”
Darrin York, distinguished professor and Henry Rutgers University Term Chair, Courtesy photo
“The key to the proposal, for me, was having a broad vision of what could be — if the right opportunity came along and if one could assemble the right team,” said York, a professor at Rutgers University in the Department of Chemistry and Chemical Biology.
York said he and other team members had been working toward that broad vision, although their efforts were independent and focused on different components. This new grant allows them to combine forces.
“We were waiting for the opportunity and resources to make the big picture a reality for ourselves and the rest of the world,” York said.
The big picture
The goal is to create an Amber package that is adaptable to all the ways computational chemists are working so they can intuitively put the tool to work on whatever problems they’re solving.
This requires having a team with a clear understanding of the chemical challenges and the technological ones. That’s what drew Aktulga, an associate professor in the MSU Department of Computer Science and Engineering, to the team and what brought him to MSU in 2014.
“At the time, I saw that the university was investing in ICER, the Institute for Cyber-Enabled Research, which told me there was going to be a lot of commitment to boosting hardware and research,” he said. “And there would be lots of opportunities for me to collaborate.”
Since then, Aktulga has worked with nuclear physicists, electrical engineers and, of course, chemists.
“They’re people with problems that need solutions. We can help deliver those with algorithm and software development,” Aktulga said. “It’s challenging, but it’s also very exciting. My collaborators do an excellent job of presenting their work understandably to computational scientists. We can understand their computational needs and what they’re trying to accomplish.”
The team will leverage the talents and technology available at MSU and ICER along with Rutgers and UC San Diego. Together, the cutting-edge computational expertise and cyberinfrastructure will enable software testing. The San Diego Supercomputer Center, or SDSC, has an NSF-funded national supercomputer Expanse, which is used by thousands of researchers for their simulations.
“We can make sure that the software we develop works well and has good performance on Expanse and other supercomputers, including new exascale machines with heterogeneous architectures that contain the latest CPUs, GPUs and other accelerator hardware,” said Goetz, assistant research scientist and group leader for Data-Driven and High-Performance Computational Chemistry at SDSC.
“This collaboration with complementary expertise from quantum mechanics to machine learning is really exciting because it puts us into a position where we can take our individual efforts to another level,” said Goetz, who is a chemist by training and has been working for years with Amber. “We want to push the boundaries of computational chemistry and provide scientists with the computational tools they need to develop solutions for problems that are out of reach today.”
The collaborative nature of the program, along with the level of funding from NSF, will also help keep talented students and postdoctoral researchers working in an academic setting, Merz added. Recently, developments in scientific computation have been largely driven by industry, including by big names such as Google.
That’s not a critique of industry — or the researchers who get jobs there — but rather a reflection of how well companies are equipped to attract talent and tackle big problems in this space.
Having developed Amber in academia for decades, Merz and York know the value of research experiences for diverse, energetic students and postdoctoral researchers in a learning environment. Merz would like to see more opportunities at universities, given their track record of innovation, and the team is thankful for this grant.
“This is an example of something that allows us to have a bigger software vision in an academic setting,” Merz said. “These grants are really, really hard to get and they’re really, really important.”
Banner image: A Spartan-led collaboration, headed by MSU chemist Kenneth Merz, is modernizing software and computational methods for next-gen hardware to help accelerate drug discovery, materials development and more. The team is working to ensure a powerful software tool known as Amber (shown above) is optimized for the high-performance computers of today and tomorrow. Credit: The Institute for Cyber-Enabled Research and Şölen Ekesan/The York Lab/Rutgers University
