By Rob Mitchum // November 20, 2012
At the molecular level, science is often a series of snapshots. With the most advanced imaging techniques, researchers can magnify targets over a million times, allowing them to examine structures as small as an Ångstrom. But in order to achieve this incredible resolution, most techniques require their targets to be fixed in place, reducing the dramatic, flowing motions of molecules to a series of before and after pictures.
To fill in the gaps, computational scientists such as those at the Center for Multiscale Theory and Simulation (CMTS), develop models that use these still images and the laws of physics to predict the movement of a molecule from point A to point B. In the case of a virus such as HIV, filling in those blanks could reveal potential weaknesses to exploit as new drug targets. In a new paper for Biophysical Journal, CMTS researchers John Grime and Gregory Voth simulated the intermediate steps of a critical moment for HIV: when it assembles a “suit of armor” for its genes.
Viruses protect their genetic information by wrapping the nucleic acids in a protective shell called a capsid. As a virus matures, the self-assembly of this capsid is a crucial step in the virion becoming infectious, making it a promising target for drug treatments that interrupt this process.
“We’re interested in the biomedical elements of this, particularly because we don’t have a lot of information that we might want to have for HIV,” said Grime, a post-doctoral researcher in Voth’s group. “If we have some idea about how the capsid is forming in the first place, and if we could stop the capsid from forming, therapeutically that’s going to be interesting.”
But how this cone-shaped capsid self-assembles is a classic example of the snapshot problem, as experiments have only revealed images of the viral structure before and after the construction. So Grime and Voth set out to build a computational molecular dynamics model that could predict the likely intermediate steps.
Current molecular dynamics simulations are capable of tracking the movement of individual atoms within a molecule over incredibly small units of time. But the complexity of tracking the motion of millions or billions of atoms simultaneously can challenge even the world’s most powerful supercomputers. So compromises must be made to find the right mix of computational feasibility and scientific utility.
“We know that at the molecular level there is vast overkill in the information we have,” said Voth, CMTS director, professor of chemistry at University of Chicago, and Computation Institute Senior Fellow and faculty. “We don’t need to know what every atom is doing at every femtosecond. So you have this trade-off in doing enough of a simplification that you don’t lose key physics, but at the same time ramp up your ability to do computations of real molecules, not models. You can really find a sweet spot there.”
Grime and Voth used this principle to develop what’s called a coarse-grained model, one that tracks elements larger than individual atoms to save on computational demand.
As a first test of the model, the researchers found they could reproduce the complex self-assembly of the HIV capsid’s walls, with an original pool of loose capsid proteins spontaneously forming into a densely packed lattice. The simulation was then used to look deeper than ever before at the basic architecture of the cone-shaped HIV capsid. Previously, scientists observed that the capsid’s lattice contains small pentagon and hexagon structures. But the new model could test whether these 5- or 6-sided pieces were the primary building blocks of the lattice, or whether there was an even smaller shape that formed earlier in the self assembly process.
Because each individual capsid protein in the model could be followed from its independent starting point to its final place in the lattice, Grime and Voth detected an earlier transitional structure: a triangular “trimer of dimers.” Those triangles then went on to form the hexagons and pentagons eventually seen in the final lattice.
“I think these triangles are the key,” Grime said. “We see them occurring in the simulations before we see things like the hexagons and the pentagons, and we see lots of them forming early. Our argument is that this shape is actually crucial to formation of the lattice in the first place; indeed, it may be the fundamental step that seeds the lattice growth.”
The coarse-grained model also allowed the researchers to test that these triangular trimers-of-dimers weren’t just an accident within the simulation. Because of the model’s reasonable size and the computational power available through the XSEDE network, it could be run hundreds of times with slightly different starting conditions to confirm that this phenomenon was indeed common and repeatable.
“That lets us check whether the effects we’re seeing are just a fluke of the way we started, or whether there is something reliable, an emergent behavior that is actually a property of the system in general,” Grime said.
Building from this first successful demonstration of their model’s ability, Grime and Voth are now looking at how the two-dimensional capsid lattice forms into the three-dimensional cone shape, while continuing to refine the model to make it more efficient and accurate. Eventually, they have the ambitious goal of creating an entire virtual HIV particle that can fully recreate the virus’ maturation and other processes…without crashing a supercomputer.
“With the sort of techniques and software we’re developing here, the idea is that eventually we will be able to model the full process from beginning to end,” Grime said. “We hope to someday actually have a full computational HIV virion, from immaturity to maturity. That that would be a very impressive demonstration of the value of computational modeling.”