10.1021/ja068336h.s001 Hans-Jürgen Sass Hans-Jürgen Sass Franziska Fang-Fang Schmid Franziska Fang-Fang Schmid Stephan Grzesiek Stephan Grzesiek Correlation of Protein Structure and Dynamics to Scalar Couplings across Hydrogen Bonds American Chemical Society 2007 water models protein G h 3JNC values h 3JNC h 3JNC data SMN Tudor domain MD simulations hydrogen bonds backbone hydrogen bonds GB 1 domain 2007-05-09 00:00:00 Journal contribution https://acs.figshare.com/articles/journal_contribution/Correlation_of_Protein_Structure_and_Dynamics_to_Scalar_Couplings_across_Hydrogen_Bonds/3008665 NMR-observable scalar couplings across hydrogen bonds in nucleic acids and proteins present a quantitative measure for the geometry and  by the implicit experimental time averaging  dynamics of hydrogen bonds. We have carried out in-depth molecular dynamics (MD) simulations with various force fields on three proteins:  ubiquitin, the GB1 domain of protein G, and the SMN Tudor domain, for which experimental <sup>h3</sup><i>J</i><sub>NC</sub><sub>‘</sub> scalar couplings of backbone hydrogen bonds and various high-resolution X-ray structures are available. Theoretical average values for <sup>h3</sup><i>J</i><sub>NC</sub><sub>‘</sub> were calculated from the snapshots of these MD simulations either by density functional theory or by a geometric parametrization (Barfield, M. <i>J. Am. Chem. Soc.</i> <b>2002</b>, <i>124</i>, 4158−4168). No significant difference was found between the two methods. The results indicate that time-averaging using explicit water solvation in the MD simulations improves significantly the agreement between experimental and theoretical values for the lower resolution structures ubiquitin (1.8 Å), Tudor domain (1.8 Å), and protein G (2.1 Å). Only marginal improvement is found for the high-resolution structure (1.1 Å) of protein G. Hence, experimental <sup>h3</sup><i>J</i><sub>NC</sub><sub>‘</sub> values are compatible with a static, high-resolution structural model. The MD averaging of the low-resolution structures moves the averages of the <i>r</i><sub>HO</sub> distance and the H···OC angle θ closer to their respective values in the high-resolution structures, thereby improving the agreement using experimental <sup>h3</sup><i>J</i><sub>NC</sub><sub>‘</sub> data. In contrast, MD averaging with implicit water models deteriorates the agreement with experiment for all proteins. The differing behavior can be explained by an artifactual lengthening of H-bonds caused by the implicit water models.