Abstract
Conformational flexibility of molecules in solution implies that different conformers contribute to the NMR signal. This may lead to internal inconsistencies in the 2D NOE-derived interproton distance restraints and to conflict with scalar coupling-based torsion angle restraints. Such inconsistencies have been revealed and analyzed for the DNA octamer GTATAATG.CATATTAC, containing the Pribnow box consensus sequence. A number of subsets of distance restraints were constructed and used in the restrained Monte Carlo refinement of different double-helical conformers. The probabilities of conformers were then calculated by a quadratic programming algorithm, minimizing a relaxation rate-base residual index. The calculated distribution of conformers agrees with the experimental NOE data as an ensemble better than any single structure. A comparison with the results of this procedure, which we term PARSE (Probability Assessment via Relaxation rates of a Structural Ensemble), to an alternative method to generate solution ensembles showed, however, that the detailed multi-conformational description of solution DNA structure remains ambiguous at this stage. Nevertheless, some ensemble properties can be deduced with confidence, the most prominent being a distribution of sugar puckers with minor populations in the N-region and major populations in the S-region. Importantly, such a distribution is in accord with the analysis of independent experimental data--deoxyribose proton-proton scalar coupling constants.