NEW methods for automated NOE assignment and NMR structure calculation [Taschenbuch]

This book describes the further development of the ARIA (Ambiguous Restraints for Iterative Assignment) method that integrates automated NOE assignment into structure calculation. ARIA starts with the calculation of an extended chain conformation which serves as a template structure for the first round of calibration and assignment. The NOEs from different spectra are calibrated using the reference distances of the template structure. They are then merged in order to obtain one list with ambiguous and unambiguous distance restraints. In an iterative fashion structures are calculated with a molecular dynamics based simulated annealing (MDSA) protocol which serve as template structures for the next iteration. Other restraints (i.e. hydrogen bonds, j-couplings, dihedral restraints or residual dipolar couplings) can be added easily. The criteria for the assignment of NOEs and for the removal of noise peaks are tightened during later iterations. As a result, more and more NOEs get unambiguously assigned. After the last iteration, the program provides a list containing all the assignments and the structures calculated with this restraint list. This allows one to check the assignments manually against the spectra. The restraints that need to be rejected during the calculation can be viewed as an integral part of the result, can be analysed in a systematic way, and might even be submitted with the structures and the active restraints. ARIA leads to a substantial speed-up of structure calculation by automation of one of the most time-consuming steps. Compared with a manual approach where initial structures are calculated based on a small fraction of the NOEs, the automated approach uses much more data to direct the calculation from the start. However, full automation has not been reached at the present state of art. In order to obtain high-resolution NMR structures, the spectroscopist still has to check the assignments. Obvious errors have to be corrected, new assignments can be added manually. The internal book-keeping of ARIA facilitates this task by providing all the information from the original peak lists together with the new assignments. While the aim is to achieve complete automation of the structure calculation, the design philosophy of ARIA is always to take advantage of any manual assignments of resonances and NOEs that are available. Thus, structure calculations with ARIA can be run either fully automatically, semi-automatically, or completely manually.

The effects of different non-bonded parameters of force fields for NMR structure calculation on the quality of the resulting NMR solution structures were investigated using interleukin-4 as a model system. NMR structure ensembles were calculated with an ab initio protocol using torsion angle dynamics. The calculations were repeated with five different non-bonded energy functions and parameters. The structure ensembles were compared with the available X-ray structures, and their quality was assessed with common structure validation programs. It was shown that structures obtained with the non-bonded energy function of PROLSQ achieve a higher quality than previous non-bonded representations. Furthermore, the results clearly indicate that larger vdW radii than commonly employed in virtually all NMR structure calculation programs significantly improve the quality of the structures, if appropriate corrections for 1-4 interactions are employed. The further improvement of the quality and the accuracy by refinement in explicit solvent with a hybrid CSDX/OPLS force field was demonstrated. The water refinement is carried out with covalent interactions identical to the structure generation. The refinement therefore does not lead to distortions in the covalent geometry.

A new CPU-efficient algorithm for the calculation of NOE intensities from structural data has been presented. The calculated NOE intensities are used for evaluating the contribution of each assignment possibility. Spin diffusion corrected distances allow the choice of smaller error bounds. The algorithm was tested with the PH domain of Mus musculus beta-spectrin. The results show clear improvements in the quality and accuracy of the calculated structure ensembles.