Prediction of proton chemical shifts in RNA - Their use in structure refinement and validation

An analysis is presented of experimental versus calculated chemical shifts of the non-exchangeable protons for 28 RNA structures deposited in the Prot ein Data Bank, covering a wide range of structural building blocks. We have used existing models for ring-current and magnetic-anisotropy contribution s to calculate the proton chemical shifts from the structures. Two differen t parameter sets were tried: (i) parameters derived by Ribas-Prado and Gies sner-Prettre (GP set) [(1981) J. Mol. Struct., 76, 81-92.]; (ii) parameters derived by Case [(1995) J. Biomol. NMR, 6, 341-346]. Both sets lead to sim ilar results. The detailed analysis was carried using the GP set. The root- mean-square-deviation between the predicted and observed chemical shifts of the complete database is 0.16 ppm with a Pearson correlation coefficient o f 0.79. For protons in the usually well-defined A-helix environment these n umbers are, 0.08 ppm and 0.96, respectively. As a result of this good corre spondence, a reliable analysis could be made of the structural dependencies of the H-1 chemical shifts revealing their physical origin. For example, a down-field shift of either H2' or H3' or both indicates a high-syn/syn chi -angle. In an A-helix it is essentially the 5'-neighbor that affects the c hemical shifts of H5, H6 and H8 protons. The H5, H6 and H8 resonances can t herefore be assigned in an A-helix on the basis of their observed chemical shifts. In general, the chemical shifts were found to be quite sensitive to structural changes. We therefore propose that a comparison between calcula ted and observed H-1 chemical shifts is a good tool for validation and refi nement of structures derived from NOEs and J-couplings.