Let's assume that at this stage you have
- Nearly complete resonance assignments
- Unassigned NOESY peaklist (possibly automatically picked)
- Converged automatically calculated structure (for example, using CYANA)
Subsequent structure refinement involves:
- Resonance assignment verification and completion
- Search for missing assignments based on initial structure (predicted short distance) and NOESY peaks that could not be assigned during automated structure calculation
- Slowly exchanging hydroxyl protons of Ser, Thr and Tyr, as well as thiol protons of Cys can be identified in the same manner.
- Peak list optimization
- Pick missed NOE peaks and adjust peak positions for existing peaks based on the initial structure. This is especially important for automatically generated peaklists. In CARA, the convenient way is to load short distances as 'SpinLinks' and display them along with peaklists for guidance.
- Remove any peaks caused by spectral artifacts, such as noise ridges, axial peaks, solvent lines, truncation wiggles, etc.
- Search output peaklists (such as '*-cycle7.peaks' from CYANA) for strong unassigned NOE peaks. Unless caused by obvious spectral artifacts, these most likely indicate incorrect or missing resonance assignments. Other causing factors could be convergence to an incorrect initial structure (more common for dimers), and exchange peaks if multiple conformations are present.
- Revise peak integration and/or calibration constants, if necessary.
- TALOS verification
- When using TALOS dihedral angle constraints verify TALOS prediction against the initial structure using the Rama viewer.
- Refinement using orientational constraints from RDCs
Optimized resonance assignments and NOE peaklists can then be used in additional rounds of automated structure calculation.
Recommendations for refinement of NOE peak assignments
- 'Consensus' peaklists - verify cases of assignment disagreement
- Temporarily tighten NOE calibration and run calculation with fixed unambiguous assignments (option 'KEEP' in noeassign). The strongest NOE violations would correspond to peaks with assignment issues. If violations are reported only for otherwise legitimate peaks, the calibration constants are too tight.