Pol is an excellent model for polymerases as a whole. It has the same overall architecture as other polymerases with related functions, but its relatively small size makes it easier to model. Accordingly, much is already known about the nature of pol . X-ray crystallography studies provide static views of pol at the atomic scale and provide experimental anchors for the models. Scientists also have some knowledge of the timing of certain events in the process—they know that the opening and closing motions surround the chemical connecting events. Researchers even have a pretty good feel for the jobs carried out by different parts of the enzyme hand. What's lacking is a detailed understanding of pol 's machinations as it goes from open to closed.

That's what the team from NYU and NIEHS want to provide—a dynamic bridge between structural views and functional data.

To model pol , the team used the CHARMM molecular mechanics and dynamics software package to make two computational representations of the polymerase in its open and closed states. The representations were created using Samuel Wilson's x-ray crystallography data and were modeled in a solution of water and sodium and chloride ions to simulate real-world conditions. A third, intermediary representation was also built by averaging the conformations of the polymerase in the first two states.

The estimated biological timeframe for the entire enzymatic process is much beyond the capabilities of current computations, even on the most advanced computers with the fastest algorithms. Numerous trajectories of the enzyme as a whole at different stages in the opening motion had to be explored, starting from the constructed models of the closed and intermediate forms. The five-nanosecond simulations used one, two, or 150 femtosecond timesteps depending on the range of the molecular force being modeled. The force calculations were then carefully merged and updated using advanced integration algorithms.

Applying these algorithms for the first time on such a large, complex biological system, the team meticulously verified the algorithmic protocols' reliability, comparing the results to those obtained through standard methods. The team's use of the integration algorithms sped up the modeling process significantly.

The simulations also greatly benefited from parallelization on the Origin2000 using the Parallel Virtual Machine protocol and Message Passing Interface on 16 processors. Still, each five-nanosecond simulation of the system of nearly 50,000 atoms required about 25,000 CPU hours.