Rotational velocity rescaling of molecular dynamics trajectories for direct prediction of protein NMR relaxation

Rotational velocity rescaling (RVR) enables 15N relaxation data for the anisotropically tumbling B3 domain of Protein G (GB3) to be accurately predicted from 1 μs of constant energy molecular dynamics simulation without recourse to any system-specific adjustable parameters. Superposition of adjacent trajectory frames yields the unique rotation axis and angle of rotation that characterizes each transformation. By proportionally scaling the rotation angles relating each consecutive pair of frames, the rotational diffusion in the RVR-MD trajectory was adjusted to correct for the elevated self-diffusion rate of TIP3P water. 15N T1 and T2 values for 32 residues in the regular secondary structures of GB3 were predicted with an rms deviation of 2.2%, modestly larger than the estimated experimental uncertainties. Residue-specific chemical shift anisotropy (CSA) values reported from isotropic solution, liquid crystal and microcrystalline solid measurements less accurately predict GB3 relaxation than does applying a constant CSA value, potentially indicating structure-dependent correlated variations in 1H15N bond length and 15N CSA. By circumventing the quasi-static analysis of NMR order parameters often applied in MD studies, a more direct test is provided for assessing the accuracy with which molecular simulations predict protein motion in the ps-ns timeframe. Since no assumption of separability between global tumbling and internal motion is required, utility in analyzing simulations of mobility in disordered protein segments is anticipated.

Highlights►Rescaling of rotational velocity in molecular dynamics for the B3 domain of Protein G reproduces its anisotropic tumbling. ►The optimal rotational velocity rescaling factor closely matches that predicted from the self-diffusion rate of TIP3P water. ►15N relaxation data of GB3 can be predicted from the rescaled trajectories with no system-specific adjustable parameters. ►15N chemical shift anisotropy of − 168 ppm predicts the field-dependent data better than reported residue-specific values. ►Sites for which observed and predicted relaxation markedly differ are assessed in terms of implied force field inadequacies.