Improving the accuracy of NMR structures of RNA by means of conformational database potentials of mean force as assessed by complete dipolar coupling cross-validation.

The description of the nonbonded contact terms used in simulated annealing refinement can have a major impact on nucleic acid structures generated from NMR data. Using complete dipolar coupling cross-validation, we demonstrate that substantial improvements in coordinate accuracy of NMR structures of RNA can be obtained by making use of two conformational database potentials of mean force: a nucleic acid torsion angle database potential consisting of various multidimensional torsion angle correlations; and an RNA specific base-base positioning potential that provides a simple geometric, statistically based, description of sequential and nonsequential base-base interactions. The former is based on 416 nucleic acid crystal structures solved at a resolution of </=2 A and an R-factor </=25%; the latter is based on 131 RNA crystal structures solved at a resolution of </=3 A and an R-factor of </=25%, and includes both the large and small subunits of the ribosome. The application of these two database potentials is illustrated for the structure refinement of an RNA aptamer/theophylline complex for which extensive NOE and residual dipolar coupling data have been measured in solution.