Chemical shift calculations

SHIFTCALC, April 2004

Mike Williamson & Mohamed Refaee
Department of Molecular Biology and Biotechnology, Sheffield UK
NMR Group

1. About SHIFTCALC

There are two main ways of calculating shifts in proteins. One is to use quantum chemical (density functional theory, DFT) methods. This is the 'purest' way to do it: you need a good model of the protein, a good algorithm, and lots of computer time. The drawback is that calculating a complete protein is still way beyond the reach of the fastest computer, so you need a very simplified model: ideally just an amino acid with a few atoms at each end, and maybe some nearby groups for hydrogen bonding. There are now some very good calculations around (see in particular the program SHIFTS by David Case's group, which is based on these kinds of calculation, but adds in some empirical results too).
The second method is to use empirical correlations. Essentially you identify some equations that look as if they might be useful to calculate shifts, based on for example backbone dihedral angles, distances to charged groups etc, and fit parameters to them until they produce calculated shifts that match the experimental ones. Typically this requires the compilation of a large database of experimental shifts together with a range of structural information. This second approach is the one we have taken, for both 1H and 13C shifts. In principle it is of course less good than the quantum chemical approach. In practice the results are similar (for Ca and Cb at least: results for CO and 15N are generally poor), and they have the distinct advantage that it is easy to understand what structural features are contributing to the calculated shift. We have therefore persevered with this method, even though at some stage quantum chemical methods are bound to replace it.
This package calculates shifts for all 1H and for Ca and Cb shifts. 1H shifts are based on [1,2,3], and 13C shifts on [4]. These have been updated, to include various more recent developments, for example the option to calculate ring current shifts for carbon [5], checking for cystine or cysteine[6], and improved random coil shifts [7]. The package is run interactively from the web, or you can download the Fortran code (see below). Any problems please email to m.williamson@sheffield.ac.uk.
References
  • 1. M. P. Williamson, T. Asakura, E.Nakamura and M. Demura J. Biomol. NMR 1992, 2 83-98.
  • 2. M. P. Williamson and T. Asakura J. Magn. Reson. 1993, 101B, 63-71.
  • 3. T. Asakura, K. Taoka, M. Demura and M. P. Williamson J. Biomol. NMR 1995, 6, 227-236.
  • 4. M. Iwadate, T. Asakura and M. P. Williamson J. Biomol. NMR 1999, 13, 199-211.
  • 5. L. Blanchard, C. N. Hunter and M. P. Williamson J. Biomol. NMR 1997, 9, 389-395.
  • 6. D. Sharma and K. Rajarathnam J. Biomol. NMR 2000, 18, 165-171.
  • 7. H. Zhang, S. Neal and D. S. Wishart J. Biomol NMR 2003, 25, 173-195.

    2. Download the source code.

    You can download programs to calculate carbon-13 alpha and beta

    and proton shifts

    The carbon-13 file is a tar file. Unpack it using tar -xvf carbon_13_alpha_beta.tar, and then follow the instructions in the read_me file.

    The programs are written in fortran and require a fortran compiler. For the shift calculation programs, it is better to compile so that reals are treated as real*8 numbers, eg with the f77 compiler use f77 -r8 total.f -o total