NMR Structures of Proteins Using Stereospecific Assignments and Relaxation Matrix Refinement in a Hybrid Method of Distance Geometry and Simulated Annealing

Abstract

The hybrid method combining the early stages of a distance geometry program with molecular dynamics/simulated annealing in the presence of NMR constraints was optimized to obtain structures consistent with the observed NMR data. Two novel methods of stereospecific assignments of the protons at the prochiral carbons are used in simulated annealing, the “floating” chirality assignment and a high-dimensional potential. These two methods were compared with stereospecific assignments obtained from the coupling constant data. There is good agreement between the three methods in predicting the same stereospecific assignments. As the high-dimensional potential uses more relaxed absolute distance constraints and also takes into account the relative distance constraint patterns, it reduces possible overinterpretation of the NOE data. The structures obtained from the hybrid method were further refined using the relaxation matrix approach. This approach employs the analytical form of the gradient of the calculated spectrum. Compared to the structures determined with the two-spin approximation, the fit to the NMR data improves significantly with only minimal r.m.s. shifts in the structure during simple conjugate gradient minimization. The R-factors, defined similarly to the crystallographic R-factors, are 0.51 for the structures calculated using the two-spin approximation and 0.26 for the refined structures. Large shifts of approx. 1 Å occur during a dynamics/simulated annealing calculation. The various stages of refinement and stereospecific assignments are tested on the NOE data for the small squash trypsin inhibitor, CMTI-I. In the case of CMTI-I, the last step of the refinement improved the agreement with the X-ray structure significantly.