The National Institute for Computational Sciences

University of Tennessee supercomputer helps bridge the gap between biological systems

Researchers use Kraken to better connect simulation of biophysical structures to direct experimentation

by Caitlin Rockett

Students of biology are taught that cells are the smallest living parts of any organism. The field of chemistry teaches students about the structure of matter at the atomic level, but often misses the connection with biology. Little if anything is regularly mentioned of the gaps in physical size and time needed to perform life-sustaining functions that exist between the atoms of a cell and the cellular organelles those atoms create.

Using the world’s fastest supercomputer managed by academia, a Cray XT5 known as Kraken located at the National Institute for Computational Sciences (NICS), a group of scientists has performed simulations that bridge the gap between length scales and timescales within cellular-sized biological assemblies.

“Essentially we’re asking, ‘How do atomic-scale interactions ultimately result in a functioning biological assembly?’” explained Gary Ayton, an assistant research professor in the Voth Group. The group, led by Gregory A. Voth of the University of Utah, focuses primarily on biological systems.

Voth’s team received funding from the National Institutes of Health to refine current computational methodology by which biological systems are simulated and to better connect simulation to direct experimental results. Access to a high-performance supercomputer such as Kraken was essential to achievement of the team’s objectives. Managed by the University of Tennessee for the National Science Foundation (NSF), Kraken has a peak performance of 606 teraflops, or 606 trillion calculations per second. Voth’s group was allocated time on Kraken through NSF’s TeraGrid, a network of supercomputers across the country that represents the world’s largest computational platform for open scientific research.

As the length scales of a system increase, so do its intrinsic timescales. Though not strictly defined, there can be four length scales within biological systems: macroscale, mesoscale, nanoscale, and atomic scale. Macroscale systems can be meters in length, their intrinsic processes operating in terms of minutes to hours, while atomic-scale systems are measured in picometers and picoseconds (one picosecond is a trillionth of a second).

Voth’s research spans the atomic level to the mesoscale -roughly the scales required to describe, for example, the cell membrane and organelles contained within it. It should be noted that the term mesoscale represents a broad range of length scales and timescales in the gamut of scientific disciplines. In geographic terms a mesoscopic system could include a mountain range, while the research conducted by Voth’s team spans only about 100 to 500 nanometers.

Simulating biological structures

Scientists have been attempting to simulate the interactions of particles in laboratories since the 1950s, but the advent of high-performance computing has revolutionized the way they can study atoms.

One contemporary method is molecular dynamics, a form of computer simulation that involves allowing atoms to interact with one another over a given amount of time under Newton’s equations of motion. Ultimately the interactions of atoms determine the properties of any given material; thus, in theory, it is possible to describe the properties of any material using quantum mechanics. It becomes impossible, however, to use only quantum mechanics in a system with a very large number of atoms, so scientists use molecular-dynamics simulations to model atoms as if they were conventional objects. Large biological molecules, such as proteins and lipids, can be modeled in this way. However, for entire biological assemblies (e.g., an entire cell), the computational limits are again reached, and it is necessary to redefine the system with “pseudo atoms,” designed to model collections of atoms. This type of simulation is called coarse graining and has been used with success in the field of biomolecular simulation.

While coarse graining can save computational time, the resolution of the models it produces is lower than that from a molecular-dynamics simulation. The current challenge is to design coarse-grain models that still contain some atomic-level “memory,” thereby increasing the predictive power of the methodology when applied to larger length scales and timescales. This necessity led Ayton to develop a code that could couple high- and low-resolution models and allow them to accurately transfer data between each other -a bridge between scales of length and time.

Ayton describes the code, TANTALUS, as a flexible, massively parallel molecular-dynamics engine of sorts designed to work both inside and outside the atomic scale.

“Most molecular-dynamics mechanisms out there expect an atomistic system to contain hydrogen, carbon, nitrogen, and phosphorous, but TANTALUS can take into it any potential function that I want to design,” said Ayton of the flexible nature of his code. “We needed a code that could apply to a broad range of length scales and timescales. TANTALUS allows us to do simulations where there are no atoms anymore.”

BAR domains and the membrane

Cellular components are kept within the cell by a fluid bilayer of lipids. This is true of all living organisms and even some viruses. This membrane functions to keep ions, proteins, and other molecules where they are needed within the cell. Occasionally, though, it is necessary for a cell to allow flow through its membrane to regulate internal conditions, which requires the membrane to temporarily change its shape.

Recent experiments have suggested that cellular membrane shape is most often determined by proteins. One of the best-studied families of membrane-sculpting proteins is the BAR domains, which are found naturally in living organisms. BAR domains are protein modules, or small parts of a larger protein, that were central to Voth’s work on membrane remodeling, an important focus of his team’s research. They have a characteristic banana shape with a positively charged concave surface that interacts with the negatively charged surface of a cell membrane, but whether these BAR domains induce membrane molding or simply stabilize it has been an ongoing debate among biochemists.

The researchers found that sometimes BAR domains act together in what Ayton refers to as the collective effect. Independent BAR domains, because of their elongated shape, along with specific interactions on their surface can sometimes line up side by side to form larger structures capable of generating larger curvatures. After one protein module has attached itself to the membrane and generated a curve, any other BAR domains in the vicinity can talk to this attached protein module. The newly attached BAR domain seeks out a place for itself on the membrane according to where it finds the lowest point of energy, which happens to be directly beside the original domain.

The collective effect is stronger than a single BAR domain, but it takes time for the domains to line up, and the process is not as orderly as it sounds. Ayton explained that one region of the membrane can have BAR domains lined up in one direction, while another region has them lined up in the opposite direction. These surfaces will not remodel, and the BAR domains have to realign as many times as it takes to get them all facing the same direction.

The team has most recently begun applying its multiscale method to the HIV-1 viron, something no other team of researchers has attempted. TANTALUS allows the team to study this precursor to the AIDS virus in a way that would be difficult or impossible in a lab.

The Voth Group’s novel approach to biophysical simulation has implications that outstretch the simulation community and flow into the realm of public health.

Of the team’s ultimate goal Ayton said, “We want to take the theoretical machinery of statistical mechanics, the computational power afforded by NICS, and use that within a multiscale context to make a significant impact and contribution to the field of molecular biology on systems that have direct health-related relevance.”