NOE R factors and structural refinement using FIRM, an iterative relaxation matrix program

Abstract

A program called FIRM ( full iterative relaxation matrix) for calculating NOE intensities and for NMR structure refinement, including contributions from internal motions, is described. Simulated NOE data sets with and without random noise were constructed for models of two proteins, motilin (22 amino acids) and parvalbumin (108 amino acids). NOE R factors were computed for structures with varying degrees of similarity to the motilin and parvalburnin models. The NOE R factor for random structures ranged from 0.8 to 1.3, and the R factor for accurate structures ranged from 0.31 to 0.39. Extended chain structures of motilin and parvalbumin were refined with the simulated NOE data sets using FIRM, distance geometry, and restrained molecular dynamics. Random noise and the choice of NOE data set had little effect on refinement. The results demonstrate that low R factors are not sufficient criteria for judging the success of a structural refinement from sparse NOE data.