Tags Posts tagged with "Dima Kozakov"

Dima Kozakov

By Daniel Dunaief

They have little in common. One studies deep inside cells to understand the difference between diseased and healthy states. The other explores models that represent distant thermonuclear reactions.

What Stony Brook University’s Dima Kozakov, Professor in the Department of Applied Mathematics and Statistics, and Michael Zingale, Professor in the Department of Physics and Astronomy, share, however, is that both led teams that recently won a Department of Energy grant that will allow them to use the fastest publicly available supercomputer in the world, at DOE’s Oak Argonne and Oak Ridge National Laboratories.

Kozakov and Zingale, who are both members of the Institute for Advanced Computational Sciences, are recipients of the DOE’s grants through its Innovative and Novel Computational Impact on Theory and Experiment (INCITE) program.

“It’s a huge recognition of computation” not just at the IACS, but also for Stony Brook in general, said Robert Harrison, Director of the IACS and Professor in the Department of Mathematics & Statistics. Kozakov and Zingale are the “point persons on world-class teams [which] positions Stony Brook at the forefront of the scientific community.”

Harrison suggested that the astrophysics group at Stony Brook was already world class when he arrived a decade ago and the university has been pushing to move Stony Brook to take advantage of all the modern powerful tools for simulation and data driven discovery.

Disease states

Kozakov, who is also an affiliate of the Laufer Center for Physical and Quantitative Biology,  plans to model enormous numbers of molecular interactions to compare how they function in diseased cells with how they work in healthy cells.

He and his team will get the data on important proteins and interactions in disease compared with healthy cells from high throughput but noisy experiments and validate those computationally.

By studying diseases such as cancer, diabetes and Alzheimer’s, Kozakov plans to look for clues about what occurs at the level of the atomic structure of protein interactions, hoping such an analysis points to the creation of new types of therapies.

Kozakov will use a combination of publicly available data and information from some of his experimental collaborators to identify new targets that small molecules may alter amid a diseased state. He feels the tight integration between the theoretical and the experimental nature of the team will enhance its effectiveness.

A supercomputer “allows you to try many approaches in parallel” such as training deep learning models that require trying many options to get the best possible ones, he said.

The pilot work the team has done created the kind of momentum that increased the chance of securing funds and time through the INCITE program.

Kozakov and co-investigators including Assistant Professor Pawel Polak at Stony Brook, Professor Andrew Emili at OHSU, Associate Professor Matthew Torres at Georgia Tech and Julie Mitchell, the Director of Biosciences Division at Oak Ridge National Laboratory, were “very happy” when they learned they’d won the award. he said. “It’s good to know that people appreciate the [work] we are doing.”

Starry, starry explosion

In the meantime, Zingale’s project, called “Exascale Models of Astrophysical Thermonuclear Explosions,” was renewed for a second year in the INCITE program.

Zingale leads a team that explores two types of astrophysical thermonuclear explosions to understand these physical processes and their broader implications. The computational work is focused mostly on whether a particular model for a thermonuclear explosion is viable.

“We really want to just understand: does it explode or not?” Zingale explained. His work focuses on the explosion mechanism and on the design of algorithms that can efficiently model these explosions.

Graduate students Zhi Chen, Alexander Smith Clark, Eric Johnson, Melissa Rasmussen, and Khanak Bhargava will be working with the supercomputer in the next year, Zingale added.

“Each student is working on separate questions, both on this problem and on related problems (novae and x-ray bursts),” said Zingale. “The goals are the same — in each case, we want to produce a realistic model of the burning that takes place in these events to understand how these explosions unfold.”

Models help connect to the observations astronomers make. While the work doesn’t produce new physics, it allows researchers to gain a greater understanding of supernovae.

Numerous other groups around the world are pursuing similar simulations, which Zingale explained is favorable for the science.

“If we all get the same result using different codes and techniques, then it gives us confidence that we might be understanding what is actually taking place in nature,” he said.

The explosions Zingale is studying differ from those on Earth because they are far larger and can reach higher densities in stars, which produces elements up to iron in explosions. The tools he uses to model these explosions have “similarities to the techniques used to model chemical combustion on Earth,” he said. “We work with applied mathematicians that study terrestrial flames and can use the techniques” in the astrophysical setting.

Zingale explained that he was always interested in astronomy and computers, so this field of work serves as the bridge between the two.

For students interested in the field, Zingale added that it teaches people how to solve complex problems on computers.

“Even if you don’t stay in the field, you build skills that are transferable to industry (which is where many of my graduate students wind up),” he said. He urges people to study something they enjoy. The main code he uses is called Castro and is freely available online, which means that “anyone can look at what we’ve done and run it for themselves,” he explained.

Student opportunities

For Stony Brook graduate students, these INCITE awards offer opportunities for additional learning and career advancement.

“The excitement is infectious,” said Harrison. “The students see not just the possibility to be at the frontier of discovery and the frontier of technology [but also to have] the career opportunities that lie beyond that.”

Students trained to make effective use of these platforms of cutting-edge science are “heavily recruited, going into industry, national labs, working for the likes of Google and so on,” Harrison added.

Dima Kozakov. Photo courtesy of Stony Brook University

A high five becomes a natural celebration after a home run because the hitter and the celebratory teammate are standing on their feet and are looking directly at each other. What if gravity didn’t keep our feet on the ground and our heads in the air? We might slap a hand into a foot or a foot into an elbow, sharing a nonverbal exchange with a different meaning.

Proteins inside our bodies don’t have the same gravitational and physical limits. They can and do come together in a soup of cytoplasm, blood, plasma and other mediums. Some of the time, those exchanges, like the high fives, communicate a message in the ordinary course of life. In other circumstances, however, those protein-protein interactions can lead to diseases like cancer.

Researchers around the world have studied these interactions using a variety of tools, trying to combat signals that contribute to damaging and life-threatening conditions.

Dima Kozakov, assistant professor in the Department of Applied Mathematics and Statistics and faculty member of the Laufer Center for Physical and Quantitative Biology at Stony Brook University, has spent several years creating a general way to model the mechanical details of how two proteins interact. This tool could become useful for researchers who are studying problematic interactions.

Leading an international team of scientists, Kozakov, who is also a faculty member at the Institute for Advanced Computational Science at SBU, created a new algorithm to model protein interactions. This algorithm accelerated how to model particular protein-protein interactions to identify harmful couplings. Kozakov and his colleagues recently published their findings in the prestigious journal, Proceedings of the National Academy of Sciences.

Applications of this technology include helping to design therapeutic proteins and speeding up vaccine design. If, for example, the interaction of a pair of proteins contributes to disease, scientists may want to design some other protein that is safe for the patient that will interact with one of the proteins. This additional coupling can avoid the more harmful protein connection.

Scientists also sometimes know that two proteins interact, but they don’t know how. Proteins often have large surfaces with many potential connections. Researchers might need to know “how two bodies come together,” Kozakov said. Proteins are flexible three-dimensional objects that consist of molecules. In modeling the interactions, Kozakov can find the three-dimensional way these proteins come together.

Computational modeling is less expensive than running experiments. At this point, the computer system needs as its starting point the three-dimensional structure of the proteins. That, Kozakov said, is much easier than determining the structure of a protein complex.

The next step is to work on methods where scientists don’t need the structure but only the chemical formula, which they can find through the amino acid sequence. Kozakov and his collaborators will use the information on the structure of similar proteins to build the models. “We’re developing a methodology that will work with the models,” Kozakov said. He described his approach as “physics based,” in which he solves a statistical mechanistic problem by using an energy function that can account for different environments.

“In principal, we can modify our energy function to account for different environments,” like changes in pH, temperature or other variables that might affect how two proteins come together. Given the way Kozakov and his colleagues designed the model, it can account for all possible configurations of two almost rigid proteins coming together.

Kozakov is also in discussions with Brookhaven National Laboratory to explore the results of small-angle X-ray scattering. The benefit of this approach is that he doesn’t need proteins in a crystalline structure, which is a requirement of crystallography. While small-angle X-ray scattering provides less information than crystallography, Kozakov said he and his colleagues can develop it in combination with other techniques where it would be equivalent.

Kozakov has been developing models since 2007 or 2008 to understand these interactions. The project in his recent paper took three years to finish. The program takes 10 to 15 minutes to run on a personal computer. Before, this kind of effort required a supercomputer.

Kozakov believes there could be other applications of this technology, where scientists could model candidate protein drugs in real time to see how the drug interacts with the protein of interest. The first version of the program came out about a year and a half ago and it took the intervening time to perfect it, he said.

Born in Eastern Europe in a region that used to be part of the Soviet Union but is now on the western border of the Ukraine, Kozakov lives in Stony Brook with his wife Olga Kozakova. The couple has a six-year old son, Platon. Kozakov’s grandparents were scientists: his grandfather, Mikhail, was a university professor and his grandmother, Nina, worked at the university. He grew up surrounded by books on physics. He “had fun, digging into antiquities books” and thought the science presented an “inspiring environment.”

As for his work, Kozakov has a big picture view of his efforts. “I want to make something useful to the community and to the world,” he said. “I want to do what I can to help.”