Using core packing to assess the quality of NMR structures*

There are two main methods for determining protein structure on the atomic scale: X-ray crystallography and NMR spectroscopy. X-ray crystallography can provide the positions of all atoms in a protein, and thus little modeling is required to determine the protein's structure. NMR spectroscopy produces numerous restraints on atomic separations and angles, which can then be combined with existing force fields to determine a protein's structure.  In contrast to X-ray crystallography, there are no accepted metrics of structure quality for NMR. In addition, there is a long-standing debate about whether NMR structures for a given protein are different from those from X-ray crystallography, and whether the differences are real or artifacts of the measurements. To address this question, we have analyzed all NMR structures deposited in the Protein DataBank (PDB) that contain the restraints used to generate NMR model structures. In comparison to high-resolution X-ray crystal structures, the set of the top 25% NMR structures with the greatest number and agreement with their restraints, includes a significant number of structures that possess more densely packed cores. Some of these NMR structures with overpacked cores possess large atomic overlaps, leading to high packing densities. However, other over-packed NMR structures have similar atomic overlap energies compared to x-ray crystal structures. We also carried out unrestrained and restrained molecular dynamics simulations of several proteins to determine whether the degree of overpacking in the cores of NMR structures is related to the number of NMR restraints that are satisfied.