Edit detail for EvaluatingTheResultsOfAutomatedStructureCalculation revision 7 of 15

This page gives some tips on evaluating the results of automated structure calculation.

UNIO

Unio includes the package AtnosCandid for automatic structure calculation. AtnosCandid automatically pick peaks in NOESY spectra, assign them to resonances provided in the chemical shift list(s), converts them to constraints, and calculates structures using an external program (e.g. CYANA). The process is iterative: the structures generated in intermediate cycles are used to reduce the ambiguity in the assignment of peaks in later rounds resulting in the identification of additional constraints.

Covalent check: The first stage of AtnosCandid analysis is the covalent check which compares the chemical shifts in the prot file to the NOESY to see if there is agreement. This take 20m to an hr depending on the computing infrastructure. Only after this stage is completed does UNIO give information on whether the input is ok and everything is set up correctly.

To do the covalent check UNIO uses the sequence and the chemical shifts along with knowledge about covalent geometry (hence the name) to calculate a predicted set of crosspeaks in the NOESYs, the covalent peaklist. These peaks should be present independent of the structure due to short interproton distances in the amino acid chain.

Once this is completed, the user should check how well the predicted peaks match the NOESY. Missing peaks (especially a tower of them) can indicate misalignment with the actual position in the NOESY.

Step-by-step:

1. open one of your NOESYs (e.g. 15N NOESY) with MonoScope? and import the corresponding covalent peaklist.
2. open the peaklist navigator with 'sl'. Double-click on a peak in the list to navigate to the peak position.
3. If you end up outside the spectrum, your AtnosCandid peaklist probably has a different order of 1H dimensions than CARA uses.
4. Use the Peaks-Rotate menu and swap the two 1H dimensions.
5. Now try double-clicking on peaks. They should match up to existing crosspeak signals in the NOESY.
6. The covalent peaklist is color coded according to whether a predicted peak was found. You can see the color code next in the PeaklistNavigator? entries under the column titled "Color".

• 1=found
• 6=missing

7. Sort the list by the PeakColor? by clicking on the label "Color" at the top of the peak color column in the PeaklistNavigator?.
8. Right-click on the PeaklistNavigator and execute "Select Color Code". Select A color with good contrast to the color that the contours have in your spectrum (e.g. yellow contrasts well with the default contour colors - both positive red contours and negative green contours).
9. In the PeakListNavigator you can toggle from display of peak positions and assignments by right-clicking on any peak in the list and selecting Show Assignments.
10. To see the full text of individual columns you may need to expand them by click-dragging on the right edge of the label at the top and dragging to the right.
11. Now with the peaklist sorted by Color (see step 7) double-click on each peak that was marked missing (color = 6) and look in the spectrum. Is there a whole tower with the same peak color indicating that all peaks in the tower were not found? Are the peaks all off center from the actual peaks in the NOESY?
12. In this case you will need to adjust the position of the corresponding spins in PolyScope?. Write down the assignment of the spin from the horizontal axis and return to it later to fix it. Ofcourse you may find multiple peaks all indicating that the same assignment is offset. You only need to write it down once.

Some notes on why peaks are not found:

1. Inaccurate positioning of a chemical shift relative to the NOESY. This could be in any of the dimensions (e.g. in dimensions 1Hnoe, 1HN or 15N of a 3D 15N NOESY). If a whole NOESY tower is not found this indicates that either the 1HN or 15N is off. You can see this by double-clicking on the peak entry in the PeakNavigator? and then comparing the position of the yellow cursor to the maximum in the 1D traces.
2. Overlap of peaks. The maximum of a weak peak may not be found if it is close to a strong peak. (I.e. the maximum found by AtnosCandid? is shifted due to overlap). This can happen accidentally and will usually only effect one peak in a tower.
3. Peak is near the solvent in one dimension and therefore obliterated by the solvent artifact, or it is gone because of solvent presaturation. Again, this is expected and can be ignored.
4. Exchange broadening. Sometimes a peak is weak due to local dynamics and so an NOE (which can be much weaker than crosspeaks in the experiments used to assign the broadened signal) might not be detectable. This is also okay.

13. Other things to look for in the covalent peaklist: unassigned resonances.

Ofcourse the covalent peaklist only predicts the subset of peaks which are present no matter what the structure looks like (local short distances) but not the peaks caused by the 3D structure (long range distances), so there will be many peaks which have no cross on them. However, what is unlikely is that an entire tower of NOEs? has no peaks on it. This indicates either of two things:

1. A resonance is incorrectly positioned in your project (that would be case 12A above) which you can fix by adjusting the position of the spin in PolyScope? (not MonoScope?!)
2. a resonance is not yet assigned. You should try to sign the resonance corresponding to the NOESY tower.

1. Perhaps it is an aromatic signal that's not yet assigned. E.g. HD signals that did not yet get included in a PHE spinsystem whose backbone resonances are assigned.
2. Or it could be a NH2 signal from either ASN or GLN sidechain.
3. Or it could be a HE signal from a MET

Such systems can be assigned from NOESY spectra, by picking them in PolyScope? (Pick System and first assign the appropriate SpinSystemType?) then select the proposed Spin Labels). E.g. if you think its a MET HE/CE tower. Pick new System. Select SystemType? MET and then select labels HE and CE

Now use Propose Spin in the NOESY strip to see what spins this signal shows NOEs to. Try to identify the nieghbourhood of the spin. Then tentatively assign it.

You can also use LUA scripts like

• "ShowNoesInMolmol.lua":AllCaluaScripts? creates molmol macro displaying a strip-anchors atom and all candidates atoms corresponding to an NOE in the strip.
• "UplsToSpinLinks.lua":AllCaluaScripts? imports the UPLs? from a structure calculation and converts them into SpinLinks? visible for a selected NOESY in a CARA project.
• "ProposeAndShow.lua":AllCaluaScripts? Displays contours for proposed SpinLinks? in the transposed positions in different NOESY spectra
• "ImportLinksFromPeakList.lua":AllCaluaScripts? Creates Spinlinks in a Project from a NOESY Peaklist