We have compared structures derived from the NOESY distance restrained molecular dynamics (MD) and unrestrained MD calculations of a tandem G.A mismatched decamer deoxynucleotide duplex d(CCAAGATTGG)2 with the X-ray crystal structure. The importance of using restrained MD in water (with counterions and a box of water molecules containing the G.A mismatched decamer) or in the gas phase is evaluated. While structures derived starting from model-built coordinates and crystal coordinates using restrained MD in both water and gas-phase environments do obey the set of input restraints, global changes in the DNA observed in the gas phase occur faster than in the aqueous medium. Thus, the use of structures generated with minimal gas-phase simulation as input for further solution-phase refinement is quite reasonable given the results observed here. Although a number of significant differences exist, many features of the crystal structure are found in the NMR solution structure. Both NMR spectroscopy and X-ray crystallography describe the rather large propeller twist at the mismatch site and the bifurcated hydrogen-bonding pattern of the mismatched guanine residue. However, the crystal structure maintains a straight rod conformation in order to allow adjacent molecules in each unit cell to stack and form the crystal. Through wedging of the adjacent thymidine residues, and thus providing opposing kinks at either end, the duplex in the crystal is able to recover its global cylindrical shape. In the aqueous environment, however, no forces are present which would cause such kinking and destacking of the thymidine-thymidine residues. Thus, an overall curvature or bend in the duplex is observed. Finally, unrestrained molecular dynamics (in a 100-ps trajectory in the gas or solution phase) is unable to adequately locate either the crystal or solution NMR conformation. The structures determined from 100 ps of free dynamics depend upon the initial starting structure.