Deviation versus violation plots: A new technique for assessing the self-consistency of NMR data.

A new method has been developed to test the self-consistency of distance constraints derived from NOESY spectra. The technique is based on the premise that the further the atomic coordinates of any given structure vary from the 'correct' structure, the more NOE violations will occur in that structure. This relationship is quantified by plotting the deviation of each structure against the sum of the residual NOE violations. This type of plot is called a DVplot, which is generated by the following means: first the experimental constraints are used to generate a set of structures, then the amount of deviation and violation is quantified for each structure. The deviation of each structure is derived from the root-mean-squared deviation (rmsd) between each structure and the average structure. Violations are quantified for each structure using the new terms S1, S2, and S3. These terms measure the sum of all residual NOE violations greater than 0.1 A, 0.2 A, and 0.3 A, respectively. DVplots are used to show that for series of structures calculated from a single set of NMR constraints, there is an approximate linear correlation between the rmsd and each of the three sums, S1, S2, and S3. Furthermore, it is proposed that the x-intercepts derived from the three plots of S1, S2, and S3 will converge if the NOE constraints are selfconsistent. The new technique is applied to five different proteins using both experimental and simulated constraint sets.