CGI for Science


Computer graphics have greatly expanded the possibilities of cinema. Special effects using CGI (computer-generated imagery) today enable directors to shoot scenes that were once considered impossible or impractical, from interstellar combat to apocalyptic action sequences to fantastical digital characters that realistically interact with human actors.

In science, computer graphics are also creating sights that have never been seen before. But where movie special effects artists are realizing the vision of a screenwriter and director, scientific computer models are inspiring new discoveries by revealing a restless molecular world we cannot yet see with the naked eye.

Using computers to peer into this hidden universe was the theme of CI faculty and senior fellow Gregory Voth's Chicago Council on Science and Technology talk last week, titled Molecular Modeling: A Window to the Biochemical World. Scientists at Voth's Center for Multiscale Theory and Simulation use computers to recreate real-world physics and produce awe-inspiring, intricate images, pushing the frontiers of discovery one femtosecond and nanometer at a time.

[Some of those images, including the one above by Mijo Simunovic, were on display as a "Science as Art" gallery, which you can view in a slideshow here.]

"The computer simulation allows us to make a movie, if you will, but it's a movie describing what the laws of physics tells us," Voth said. "It's not a movie where we tell the computer we want this figure to run and shoot this figure. We don't know what's going happen. We know the equations, we feed them in [to a supercomputer], and we solve those equations…and we can reach scales we never dreamed of reaching before."

Those equations are Newton's laws of motion and quantum mechanics, applied in a molecular dynamics model to simulate the activity of individual atoms. For a single molecule of water, this would only require calculating the equation for three atoms. But to simulate a volume of thousands of water molecules, or a protein made up of millions of atoms, or thousands of those proteins, an incredible number of calculations must be done at every time step of the model. An increment, Voth said, that is usually a single femtosecond, or one-millionth of a nanosecond.

If the technical demands can be met, these simulations can unlock a flood of previously inaccessible information for chemists and biologists. While laboratory methods can produce still images of proteins in various states, molecular simulations can create living animations of the transitions between those states, or how they interact with other proteins or the surrounding environment. That additional insight allows scientists to study subjects such as the HIV virus, cell membranes, actin filaments or ATP hydrolysis in unprecedented detail.

In the CMTS, Voth and his collaborators are looking for ways to increase the complexity of these models without pushing them beyond the point of practicality on today's supercomputers. Through a method called coarse-grained modeling, proteins aren't reduced all the way to their constituent atoms, but instead to "beads" that aggregate the activity of several atoms, requiring fewer calculations and less computational power.

"What this does for you, if you do it well, is it's a bridge," Voth said. "It's an intermediate bridge between the molecular world and the mesoscopic world, and it's dramatically more efficient to solve the equation behind this to push up into the scale model and finish the data."

As an example of molecular modeling's real world potential, Voth used a project with post-doctoral scholar John Grime on HIV that was featured on ScaleOut last year. Voth and Grime are constructing a model that simulates the construction of the HIV capsid, a key step in the virus' maturation that encloses its genetic material in a protective "suit of armor." Their current coarse-grained model is not yet powerful enough to build the complete capsid, but can form "pieces of its shell" that reveal important building blocks – trimers of dimers – that are a promising target for drug makers.

Eventually, Grime and Voth hope to build a model that simulates an entire HIV virion, allowing for even more in-depth study of its behavior and weaknesses. Similarly, computational biologists hope to build models of living cells and molecular engineers are using computer models to find new ways of building nano-scale materials for energy storage and other uses. While ever more powerful supercomputers will be crucial to running these models, the people designing smarter and more accurate ways of simulating cellular processes are even more important, Voth said.

"We've come a long way. It didn't come just from faster computers," Voth said. "It came from a combination of ideas, theory and statistical mechanics. Put all that together with our powerful computers, and we can do all of these beautiful things."

Written By: 
Research Tags: