18] Error analysis of macromolecular structures determined with nuclear magnetic resonance data

Abstract

Publisher Summary This chapter focuses on the error analysis of macromolecular structures determined with nuclear magnetic resonance (NMR) data. Error analysis is restricted to a discussion of precision, which should be distinguished from accuracy. There are two basic methodologies for obtaining a structure from NMR data: one approach uses distance constraints extrapolated from the nuclear Overhauser effect (NOE) data and the other fits the NOE data directly. Precision can be approached experimentally, whereas accuracy can only be addressed theoretically because the true value can never be known through experiment. This chapter emphasizes on defining precision. The precision is typically expressed using root mean square deviations (RMSDs) of the atomic coordinates of an ensemble of structures obtained by executing repeated, independent “fits” of the same constraints. A more rigorous measure of the precision of an NMR structure can be obtained by using the Monte Carlo method for determining the precision of parameters obtained from a nonlinear least squares fit. This method has certain advantages: (1) it provides a measure of the precision, which is based on the absolute magnitudes of the NOE residuals; (2) incorporates the noise of the NOE effect spectroscopy (NOESY) data; (3) takes full account of the correlation of the coordinates, (4) is firmly rooted in accepted practice of error analysis; and (5) is free of user intervention.