Conformationally selective multidimensional chemical shift ranges in proteins from a PACSY database purged using intrinsic quality criteria

Keith J. Fritzsching # , Mei Hong and Klaus Schmidt-Rohr #
J. Biomol. NMR


We have determined refined multidimensional chemical shift ranges for intra-residue correlations (C–13C, 15N–13C, etc.) in proteins, which can be used to gain type-assignment and/or secondary-structure informa- tion from experimental NMR spectra. The chemical-shift ranges are the result of a statistical analysis of the PACSY database of [3000 proteins with 3D structures (1,200,207 13C chemical shifts and[3 million chemical shifts in total); these data were originally derived from the Biological Magnetic Resonance Data Bank. Using relatively simple non-parametric statistics to find peak maxima in the dis- tributions of helix, sheet, coil and turn chemical shifts, and without the use of limited ‘‘hand-picked’’ data sets, we show that >94 % of the 13C NMR data and almost all 15N data are quite accurately referenced and assigned, with smaller standard deviations (0.2 and 0.8 ppm, respectively) than recognized previously. On the other hand, approxi- mately 6 % of the 13C chemical shift data in the PACSY database are shown to be clearly misreferenced, mostly by ca. -2.4 ppm. The removal of the misreferenced data and other outliers by this purging by intrinsic quality crite- ria (PIQC) allows for reliable identification of secondary maxima in the two-dimensional chemical-shift distribu- tions already pre-separated by secondary structure. We demonstrate that some of these correspond to specific regions in the Ramachandran plot, including left-handed helix dihedral angles, reflect unusual hydrogen bonding, or are due to the influence of a following proline residue. With appropriate smoothing, significantly more tightly defined chemical shift ranges are obtained for each amino acid type in the different secondary structures. These chemical shift ranges, which may be defined at any sta- tistical threshold, can be used for amino-acid type assign- ment and secondary-structure analysis of chemical shifts from intra-residue cross peaks by inspection or by using a provided command-line Python script (PLUQin), which should be useful in protein structure determination. The refined chemical shift distributions are utilized in a simple quality test (SQAT) that should be applied to new protein NMR data before deposition in a databank, and they could benefit many other chemical-shift based tools.