With a well-tested, reliable equation for estimating conformation using carbon chemical shift data, researchers may some day be able to simply plug-and-chug their way from readily available NMR data to tricky conformational estimations. But the method developed by Moyna's team isn't quite there yet.

"This is a preliminary study. We are still working on proof of the method," Moyna says. To that end, the team is testing the technique used to derive the functions, making sure the estimations yielded by the functions are valid.

First, the team is confirming that they used sufficiently exacting basis sets—mathematical approximations of atomic orbitals used to solve the Schroedinger equations at the heart of any ab initio calculation. Very precise approximations yield very precise output but require outrageous amounts of computing time. In their original calculations, the team used relatively imprecise basis sets. Now they're going back and doing some of the calculations again with more time-consuming sets. They're finding that their original approximations were ample to begin with.

"You can use huge basis sets, but it can take 10 or 20 hours for a single calculation even on the newest Alliance hardware. It will be very exact, and your theoretical work will likely correlate very closely with the experimental data. But you have to find a compromise between what is computationally feasible and what still gives you good results," Moyna says.

"We went quick and dirty to begin and compared that to calculations that were much more taxing, and we found that we got near linear correlation between the two approaches. We are getting high-theory estimates on the conformations at low-theory costs."

While completing their comparisons of basis sets, the team is also looking at different ways of solving the equations used to calculate the optimal structures of the disaccharides before the chemical shifts are derived. They are also conducting large-scale molecular dynamics simulations of the disaccharides to further validate their method.

Many of these calculations are being carried out on the Alliance's new HP N-4000 complex, a 96-processor cluster of HP N-Class servers. Some of the work is also being done in-house at the University of the Sciences in Philadelphia. The platform shift hasn't bothered the team, though.

"We're using Gaussian 98 [a commercial computational chemistry code], just like everybody and their brother," Moyna says. "It was on both systems, so moving over to the new HP system was no problem. We love it."