Analysis of the peaks in the NMR spectra is obviously much more useful if one has the assignment of which atoms in the molecule (or molecules) give rise to which peaks in the spectrum. Section 7 will focus on some tools in NMRViewJ for the initial assignment of peaks from triple-resonance (13C, 15N, 1H) spectra and Section 5.7 on the more specific problem of identifying the pairs of hydrogen atoms that give rise to NOE peaks. Here we will describe some tools that are of general utility in working with the identification of NMR peaks.

Every peak list stores within itself several parameters that facilitate peak searching and matching. The peak template parameter is a list of dimension name and tolerance pairs that are used in searching peak lists for peaks with certain chemical shift values. The search template can have fewer dimensions than the peak list. So, for example, a 3D HNCO experiment could have its template set with nv_peak template hnco 1HN 0.1 15N 0.3. Searching this peak list with two chemical shifts, nv_peak find hnco 8.3 117.0, would find any peaks whose 1HN chemical shift is between 8.2 and 8.4 ppm, and whose 15N chemical shift is between 116.7 and 117.3 ppm. Because the 13C dimension is not specified in the template the 13C chemical shift is ignored in the search.

The peak-list's pattern and tolerance parameters are used when a search is done to find atoms whose assigned chemical shifts are consistent with those of a given peak. The peak pattern of an individual dimension has the format sequenceNumber.atomType. For example, a peak pattern could be set for a peak list named hnco with the command: nv_peak pattern hnco i.hn i.n i-1.c. This implies that any atom assignments for the first dimension should be of atom type hn, for the second dimension should be of type n, and the third dimension of type c. Because the first and second dimensions share the same symbolic sequence number (which can be either an i or a j the atom assignments for these two dimensions should always have the same sequence number. The third dimensions atom assignment should have a residue that is one less than the residue assignment for the first and second dimensions. The tolerance value indicates the allowable deviation between the peaks chemical shifts and the chemical shift of any atoms assigned to the peak. For example, nv_peak tolerance hnco 0.1 0.3 0.4, would set the tolerance for the first dimension to 0.1, the second dimension to 0.3, and the third to 0.4. The peak pattern and tolerance values are most readily set within the Peak Reference panel. Note: the peak pattern and tolerance values are not enforced for peaks in the present versions of NMRViewJ. They exist as storage for parameters that can be used by Tcl scripts for the analysis and assignment of peaks.

The following shows the Peak Reference panel for an 15N NOESY-HSQC experiment. The first and third dimensions are for the amide proton and nitrogen, respectively, as can be seen from the patterns "i.h" and "i.n". The second dimension is for the indirectly detected proton which could be on the same or a different residue, and is not necessarily an amide proton, so it has the pattern "j.h*". The proton in the first dimension has relation D3, this means it is bonded (the D comes from descendant, in a tree structure) to the atom in the third dimension (that is, the amide proton is bonded to the amide nitrogen). The relation field is not necessary in this case as the only atoms that would match the i.h pattern are the amide protons, but in many cases it is necessary to explicitly indicate which dimensions correspond to atoms which must be bonded to the the atom in another dimension.

The pattern and tolerance parameters are central to the use of the Peak Identification Dialog. This dialog, brought up from the Edit->Identify menu of the Peak Analysis panel, can be used to interactively assign atom identifiers to individual peaks for a molecule with existing chemical shift assignments. If this dialog is open, then each time you step to a new peak in the Peak Analysis panel this PeakID panel will be updated with a list of the atoms. The atoms listed are those atoms for which their chemical shifts are within the peaklist's tolerance of the peaks chemical shift values, and for which their atom names are consistent with the peaklist?s pattern.

Each possible assignment will be listed in the table with a set of scores indicating how close the chemical shifts of the atoms in that assignment are with the chemical shifts of the corresponding dimensions of the current peak. If three-dimensional coordinates of the atoms are available, and two of the peak dimensions correspond to hydrogen atoms, then additional information is given about the distance between the pair of hydrogen atoms. The columns of the table contain the following information (the column numbers given are for the example 3D peak list, and will differ for peak lists with a different number of dimensions).

The most probable assignments, based on the distances in the current structure set are those with the highest assignment contribution, as given in the last column of the table. The column is sorted by this value, so the most probable assignments are at top. After the table is populated a subset of the rows will be selected (highlighted in yellow in above example). This is the minimum set of assignments required for the sum of the assignment contribution to exceed a target value. This value, by default 1.0, can be set by entering a value in the CLimit field. The exponent applied to the distances when calculating the contributeions is by default 6.0, and can be set by entering a value in the Exp field.

Only those assignments where at least one structure has a distance less than a specified cutoff will be shown. By default, all assignments are shown. The cutoff can be set by entering a value in the CutOff field.

If the Spectrum1 or Spectrum2 checkboxes are activated then the region of the spectrum (of the dataset assigned to the peaklist) around the peak can be displayed in the panels on the right side of the PeakId panel. Double-clicking on an entry in the atoms list will refresh the display of the spectrum and draw the crosshair cursor at the position of the chemical shifts of the atoms in that entry.

NOE spectra are often collected so that the experimental sweep width is less, in the indirect dimensions, than the sweep width needed to include all the peak resonances. In this case the peaks will appear at a position folded by some number of sweep widths from their actual position. These peaks would not normally be matched with atoms whose chemical shifts are at their correct, unfolded, position. The peak identifier can be configured to search not only at the actual peak position, but at one or more folded chemical shifts. This can be configured using the dialog brought up by clicking Configure Folding.

The NMRViewJ user should themselves synthesize the information from the chemical shift deviations, hydrogen pair distances, and spectral display to reach a conclusion about which, if any, of the atom entries are the correct assignment for the peak. Selecting the entry and then clicking the Assign button will update the assignment labels for the current peak with the names of the atoms in that entry.

At any given point in a project it may not be possible to assign a unique set of atoms to a particular peak. Instead, because of overlap of chemical shifts there may be an ambiguous group of atoms that could give rise to the peak. NMRViewJ, supports the explicit assignment of an ambiguous set of atoms to a particular peak. This can be done within the PeakId panel by selecting more than one row of atom sets in the list and clicking the assign button. The entire group of atoms will be stored with the peak. These assignments can be viewed within the Peak Analysis panel by using the arrows adjacent to the peak label entry.

Links and Resonances

In this manual there will often be references to Links and to Resonances. Fundamentally, these refer to the same thing. Both terms are used because of the historical manner in which these were implemented in NMRViewJ, and because sometimes one or the other term is a more natural fit to the topic being described.

In an early implementation of NMRView it was found useful to be able to associate peaks with each other. For example, all the peaks that represent a given proton in a set of backbone assignment experiments (HSQC, HNCA, HNCO, ...) might be linked to each other. These links were internally maintained using the computer science concept of a bi-directional linked list so the abstract concept of links was actually implemented by computer code of the same name. Commands were written (nv_peak link, nv_peak unlink) to set, query, and remove links between peaks. Having created these links it was possible to propagate the assignments of one peak to another in the group of linked peaks. But this did not happen in any automated way, one had to explicitly invoke a command to transfer the assignments from one peak to all its partners.

During the develpment of the CCPN project, and the complemtary development of the BMRB STAR version 3 archive format the concept of a Resonance was introduced. The Resonance concept and links shared many useful features, but the Resonance concept certainly has some advantages. The above set of linked peaks can be represented using Resonances by setting all the peaks to share a single Resonance "object". The Resonance can be seen as representing the underlying atom that gives rise to each of the peaks. In fact, both Resonances and Links, actually refer to individual peak dimensions. So all the proton dimensions of a set of "linked" peaks share one Resonance, and all the nitrogen dimensions share another.

A key advantage of using the Resonance concept is that assigning a peak is actually done by assigning the corresponding Resonance. Then all peak dimensions that share that same Resonance immediately share the same assignment. There is no need to actually propagate the assignment across the set of linked peaks.

While NMRViewJ implements Resonance, the older concept of linking is still used in practice. When one links (with the "nv_peak link" command), one is actually removing the Resonance of one of the peaks, and replacing it with the Resonance of the other. The peaks now share a single Resonance, and can be considered to be "linked". Each Resonance has an identification number so you can actually tell if peaks are linked by looking at this number in the peak inspector. If they have the same number in the Resonace field, they share the same Resonance.