A simple spectral-driven procedure for the refinement of DNA structures by NMR spectroscopy.

We have developed a simple and quantitative procedure (SPEDREF) for the refinement of DNA structures using experimental two-dimensional nuclear Overhauser effect (2D NOE) data. The procedure calculates the simulated 2D NOE spectrum using the full matrix relaxation method on the basis of a molecular model. The volume of all NOE peaks is measured and compared between the experimental and the calculated spectra. The difference of the experimental and simulated volumes is minimized by a conjugated gradient procedure to adjust the interproton distances in the model. An agreement factor (analogous to the crystallographic R-factor) is used to monitor the progress of the refinement. The procedure is an The agreement is considered to be complete when several parameters, including the R-factor, the energy associated with the molecule, the local conformation (as judged by the sugar pseudorotation), and the global conformation (as judged by the helical x-displacement), are refined to their respective convergence. With the B-DNA structure of d(CGATCG) as an example, we show that DNA structure may be refined to produce calculated NOE spectra that are in excellent agreement with the experimental 2D NOE spectra. This is judged to be effective by the low R-factor of approximately 15%. Moreover, we demonstrate that not only are NOE data very powerful in providing details of the local structure but, with appropriate weighting of the NOE constraints, the global structure of the DNA double helix can also be determined, even when starting with a grossly different model. The reliability and limitations of a DNA structure as determined by NMR spectroscopy are discussed.