Peaks and Assignments

Assignments

Making Assignments

You assign peaks using the assignment dialog (at). Select a peak in a view. Bring up the assignment dialog (at) and type in group and atom names for each axis and press Apply. The peak is labeled with the assignment.

The assignment dialog contains a list of resonances for the molecule and condition of the selected peak. If you click on an axis label (w1, w2, ...) in the assignment dialog the resonance list will scroll to the resonance closest in frequency to the peak position along that axis. If you double click on a line in the resonance list it will fill in that group and atom name for the selected axis.

The Resonance Peaks button brings up a list of peaks from all spectra with an assignment involving the resonance selected in the resonance list.

Deleting Assignments

The assignment delete command (aD) deletes the assignments from all selected peaks.

Guessing Assignments

Press the Guessing button in the assignment dialog (at) to set up assignment constraints. You can consider intra-residue (i,i) guesses or inter-residue (i-1,i) guesses. You can request guesses using only a specific atom name or set of names separated by '|'. You specify the range of chemical shift values away from the peak position to look at. Press the Apply or Ok button after choosing the constraints. Now whenever you show a peak from the same spectrum in the assignment dialog Sparky will look at your current set of resonances and find the closest assignments satisfying the constraints. It puts these in a menu under the "guesses" button in the assignment dialog. Selecting a guess from this menu fills in the assignment. The guessing constraints are also used by the "assignment guessing" pointer mode (sh-F2). When you select peaks in this pointer mode they are assigned if they have a unique assignment guess. This mode will not reassign a peak. Only new assignments are made.

Assignment Tables

To display a table of chemical shifts where the rows are the different groups (ie. residues) and the columns are the different atoms use the assignment table command tb. The chemical shifts shown in the table are averages obtained from all peaks in all spectra having the same molecule and condition names. To see the number of assignments contributing to each average chemical shift click on the "assignment counts" switch and press the "update" button. You can also have the table entries include the standard deviation of the peak positions contributing to each chemical shift. To display the list of the peaks contributing to a resonance click on its entry in the table.

Copying peaks and assignments

You can copy peaks with their assignments between spectra. You select the peaks, use the ornament copy command (oc), select the spectrum window to copy to, and use ornament paste (op). For copying peaks across the diagonal of a homonuclear 2D spectrum select the peaks to be copied and use the crossdiagonal ornament copy command oX.

To copy just assignments, assuming peak markers already exist in the destination spectrum, you can use the same two step copy and paste method with the "assignment copy" (ac) and "assignment paste" (ap) commands. The assignment copy command remembers the currently selected set of selected peaks. Then you select the window for a different spectrum and assignment paste tries to apply the assignments to any peaks that are near the positions of the source peaks.

There is also an assignment copy pointer mode. In this pointer mode you just a rectangle around the peaks you want assignments applied to. The source spectrum is specified with the assignment copy dialog (co). Peaks in the source spectrum near the selected peaks are found and their assignments are copied to the selected peaks.

When doing an assignment copy the source peak and the destination peak must have nearly the same ppm position and there must not be any other peak in the source spectrum or destination spectrum that is nearby. What constitutes "close" or "nearby" is determined by the ppm shift tolerance set in the assignment copy dialog (co). Each axis of each spectrum has its ppm tolerance. When determining if there is a nearby peak marker in a given spectrum, the tolerances for that spectrum are used.

For copying assignments across the diagonal of a 2 dimensional homonuclear spectra (eg. 2D NOESY) there is the command aX. You select the peaks you want assignments applied to and assignments are found from symmetrically located peaks across the diagonal.

Chemical shift list

The resonance list command rl shows a list of resonances for a specified condition. Each line shows a group name, atom name, average chemical shift from all assigned peaks, standard deviation, and the number of assignments made from all spectra. The list can be saved to a file. The list is not automatically updated when new assignments are made or peaks are moved. Press the update button to update the list. After selecting a new molecule / condition from the menu you need to press the update button to show the resonances. The Load button allows you to read a list of chemical shifts from a file. The lines should have the same format as the first 4 columns that the resonance list displays:

	Group   Atom  Nuc    Shift

	   G15    H1'   1H    6.029
	   C17    H2"   1H    2.817
	   A18    H2'   1H    2.688

Additional columns are ignored. If a resonance in the file already has peak assignments in your spectra then reading the resonance from the file has no effect. The main purpose of the Load capability is to load chemical shifts obtained from other programs so that they can be displayed along the edges of contour plots, and used by the assignment guessing tools.

Renaming groups and atoms

Use the rename resonances (rr) dialog. If you want to change a group name fill in the old and new group names. Leave the atom names blank. To change an atom name for all groups fill in the old and new atom names and leave the group names blank. To change a specific resonance enter old and new atom and group names. Only resonances for the given molecule and condition are changed. If the condition is left blank then resonances for all conditions of the specified molecule are changed. If the condition and molecule is left blank all resonances for all molecules are renamed. In summary, leaving a field blank means apply the rename for all values of that field. You can delete resonances not used in any peak assignment by renaming them to nothing.

Predefined resonances

You can copy resonances from one molecule / condition to another. Use the "predefined resonances" command (pd). Copied resonances can be used to aid assignment of a new spectrum. They can be shown along the edges of views (vR) and appear in the list of resonances in the assignment dialog. And they are used when Sparky guesses assignments. Once a resonance is used in a peak assignment its frequency is computed from the average positions of the assigned peaks. To delete resonances that are not used in any peak assignment use "delete unused resonances" (dr). To delete only a few unused resonances, use resonance rename (rr) and rename them to nothing. If you have a chemical shift produced outside of Sparky and would like to read it in to aid in making assignments use the Load button on the resonance list dialog.

Displaying assignments on a structure using Midas

You can show assignments as colored lines between atoms on a structure using the molecular display program Midas. Midas does not come with Sparky but is available from the Computer Graphics Labs. If you install Midas in a non-standard location you will need to adjust the path in the Sparky startup script. The mi command shows a dialog where you select a spectrum, PDB file, and line color. Sparky starts Midas and displays assignments for the selected peaks. There is a switch to request that all peak assignments be shown. In this case the assignments for selected peaks will be higlighted. If you select a new peak or peaks the lines in the Midas display are automatically updated. Some assignments will not be displayed because the assignment atom name doesn't match a pdb file atom name. Sparky peak lists can display the distance between atoms in an assignment and the peak lists can be sorted by this distance. By selecting an entry in the peak list, you select the peak and the Midas display will highlight the assignment. You can only display assignments from one spectrum at a time. You can change the spectrum, line color, or pdb file in the mi dialog at any time and press Apply to change the displayed assignments. The lines representing assignments are given a Midas "model number". The lines can be displayed or hidden with the Midas commands "objdisplay model-number" and "~objdisplay model-number".

Peaks

Aliased peaks

You can mark folded peaks with the correct frequency. This allows peak lists to display correct chemical shifts. It is also needed to compute correct average chemical shifts from all assigned peaks. To say that a selected peak really belongs off spectrum shifted one sweepwidth down field along the w1 axis use a1, or up field use A1. For axis w2 use accelerators a2, A2.

Spectra acquired in TPPI mode can create aliased peaks that are reflected about the up field or down field edge of the spectrum. The u1 command gives the selected peak an alias by reflecting the peak w1 position about the up field spectrum edge. Using the u1 command a second time removes the alias from the selected peak. The u2, u3 and u4 commands mark selected peaks as aliased in the w2, w3 and w4 dimensions. The d1, d2, ... commands perform the reflection about the down field edge of the spectrum.

Centering peaks

To center a peak at the local maximum use "peak center" (pc). This uses quadratic interpolation to position the peak to a resolution finer than one data point. If you want to type in the position select a peak and use the ornament properties dialog (ot).

Deleting peak markers

To delete a peak marker select it and press the delete key. You can select peaks using the mouse in the Select pointer mode, or you can click on lines in a peak list. If you select multiple peaks, they are all deleted when you press the delete key. If you delete some peaks by mistake use the "edit undo" command eu to restore them. The eu command only undoes the effect of the most recent peak delete, peak integration or peak repositioning.

Estimating peak linewidths

The most accurate linewidths are obtained by integrating with Gaussian or Lorentzian fitting. This can be inconvenient because fitting sometimes fails. The "estimate linewidth" (pe) command gives full width at half the peak height for all selected peaks. This command first centers the peaks at the local maxima. You can manually set linewidths using the ornament properties dialog (ot). The linewidths can be seen in peak lists. Instead of looking through a peak list you can have the linewidth displayed in a status bar at the bottom of the view window when the pointer is over a peak. Use show peak info (vp) to turn this status bar on or off.

Peak groups

You can define a set of selected peaks as a peak group (pg). Peak groups have a volume and an average position weighted by the peaklet volumes (if integrated). They are intended to represent multiplet peaks.

Integration

Integration Methods

You can do two types of integration. The preferred method is to fit peaks to Gaussian or Lorentzian functions and get the volume from the fit height and linewidths. In cases where this doesn't work you can sum up the data heights at each point in a box or ellipse surrounding the peak. You choose the box or ellipse by dragging out the region with the mouse while in the "integrate" pointer mode. For the fitting method you can also use the "integrate" pointer mode and drag a region around the peaks to integrate or you can select the peaks and use the "peak integrate" command (pi). You set the method of integration using the integration dialog (it). In rare cases you might want to set the peak volume by hand. For instance, if you have a doublet peak and one component is overlapped you could integrate the unoverlapped part and manually double the volume. You type in the volume value by selecting the peak and using the ornament properties dialog (ot).

Fitting Gaussian or Lorentzian line shapes

When you integrate by Gaussian or Lorentzian fitting you need to specify the data to use for the fit. The default is to use data above the lowest displayed contour level. There is a switch in the integration settings dialog (it) to use data points within a rectangle drawn with the mouse. This is only possible for 2-D spectra. You can also select data both within a rectangle and above the lowest contour.

If the lowest contour contains more than one selected peak and the "grouping method" in the integration dialog is set to "common contour" then all the selected peaks within the contour will be fit simultaneously. That is the positions, heights and linewidths for all peaks are adjusted concurrently to achieve the best fit to the data. If you don't want peak positions or linewidths to change, switch off "Allow peak motion?" and/or "Adjust linewidths?" in the integration dialog (it). Even if peaks do not lie within the same contour they can be simultaneously fit if they are within a specfied distance of one another. You turn on "Group close peaks?" in the integration dialog and specify a maximum distance (in Hz) for each axis. Two peaks are grouped if they are within the specified distances along all axes.

If you are fitting a peak that is close to another peak that has already been fit you might want to turn on "Subtract fit peaks?". This subtracts the fits of other peaks from the data before fitting the new peak.

The "Fit baseline?" option adjusts baseline offset and slope parameters when doing the fit. For this to work well you need to use data around the base of the peak. It is convenient to set "Use data within rectangle?" on and "Use data above lowest contour?" off when fitting baselines.

If a peak has been successfully integrated, the peak marker has a circle drawn around it. The fit linewidths and volumes and residual errors can be inspected in a peak list. This information is also output to the main Sparky window when the integration is performed. To see the fit and data in 1-D cross-sections turn on slice panels. The slices do not show the baseline fit (because each fit can have a different baseline). Only the Lorentzian or Gaussian part is shown. To judge whether the baseline fit is good you can have the 1-D slices show the difference between the spectrum data and the Lorentzian or Gaussian shape (sf). What is left should be a smooth baseline. You can also have the contour plot show the spectrum data with peak fits subtracted (vf).

The fitting code uses the simplex method to adjust the positions, heights, and linewidths of all peaks within a common contour simultaneously. It minimizes the sum of squares between the Gaussian or Lorentzian line shape and the actual spectrum data at spectrum points lying withing the lowest contour or dragged rectangle. The "fit residual" is reported in the main Sparky window as rms as in the following example:

  Isolated peak @  4.796  7.797 lw 30.093 20.221 vol 9.244e+06 rms 4.5%

It is also available in peak lists (lt) by using the Options... button on the peak list dialog and switching on the "fit residual" field.

The residual for a peak is calculated as follows. Take the sum of the squares of the differences of spectrum values from the Gaussian or Lorentzian fit at each spectrum point used in the fit, divided by the number of data points and take the square root to get a root mean square deviation of the spectrum height from the fit. Multiply this times the number of data points to estimate the volume error and then divide by the fit volume to express it as a fraction. That fraction is what is shown as rms or "fit residual". It is a measure of the volume error due to the data not matching the Gaussian or Lorentzian shape.

Things that go wrong when fitting peaks

Each spectrum has its own set of integration settings. You can select the spectrum at the top of the integration settings dialog. There is currently no way to change the settings for all spectra at once. Remember that when you change a setting you must press the Apply button or the Ok button for the change to take effect.

The fit procedure uses an iterative procedure to minimize the sum of the squares of the errors between the fit curve and spectrum data. The iterative procedure can move peaks far away or give them unreasonable linewidths. To prevent this you can specify the maximum distance any peak should be moved along each axis. And you can specify upper and lower bounds for linewidths. You can also handle this by turning off "Allow peak motion?" and "Adjust linewidths?". You can copy linewidths and positions from other peaks with the same assignment that have already been integrated using the copy linewidths (cl) extension.

If you try to simultaneously fit the linewidths, positions and heights of a group of more than 5 peaks (25 degrees of freedom) be prepared to wait a while. You might hit the cancel button that appears while the fit is being calculated, raise the lowest contour level (ct) so that the group separates into smaller groups, and retry the integration. Or you can integrate by the box or ellipse method.

If the lowest contour enclosing a peak encompasses a giant region, for example, if it includes all the diagonal peaks, then the integration will fail with a message about the contour being too big. You can raise the contour level to make the region small. Or you can turn on "Use data within rectangle?". If a peak is below the lowest displayed contour (ie. has no contour surrounding it) then it will not be integrated if the fit uses only data above the lowest contour because in this case there is no data to fit to. Lower the contour levels or turn on "Use data within rectangle?".

Unintegrating

The unintegrate command pU erases the volume and linewidth of a peak. You might use this on peaks whose volume seems unreliable because of overlap. By unintegrating you can eliminate questionable volumes from peak lists.

Peak labels

You can place labels on peaks that are not assignment labels. These labels can contain any text you want. Select a peak and use the peak label command (pl). Or select the "add label" pointer mode and click on peaks to label them. A dialog will come up for you to type in the label.

To place assignment labels on all selected peaks use the lb command. To remove labels from selected use the lB command.

To automatically have labels in the current view moved so they do not overlap one another or peak markers use the unoverlap labels command lu. Each visible label is moved by small amounts in the direction (up, down, left or right) which minimizes its overlap with other labels and peak markers. This procedure is iterated to unoverlap all labels. If you are not satisfied with the results you can immediately move the labels back with the undo command eu. Note that in 3 or 4 dimensional spectra arranging the labels to be non-overlapping in one plane usually makes them overlap when viewed in other planes (because additional labels become visible). This command does not know to avoid overlapping the arrows connecting assignment labels to peaks.

The right offset labels command lr moves all visible peak labels to the right of their peak markers. The gap between the left edge of the label and the center of the peak marker is twice the peak marker size. Use the undo command eu immediately after to move the labels back to their original positions if you don't like the results.

Peak lists

Use the peak list command (lt) to bring up a list of all peaks in a spectrum. If there is already a peak list for the spectrum its window will be raised on top of all other windows. To force a second peak list window to be created for a spectrum use the new peak list command nl. Peak lists can display assignments, chemical shifts, volumes, linewidths, signal to noise ratios, data height, deviation from average resonance frequencies, .... Press the Options button to change the set of fields displayed. Use the Save button to write the list to a file. You will probably need to produce peak lists in a precise format for input to other programs that calculate molecular structures. You can either massage the peak list file using your favorite tool (Perl, awk, Tcl, vi) or you can write a Sparky extension in the Python language to directly output the desired format.

Selecting peaks

Lines in a peak list are highlighted if the peak is selected. Clicking on a peak list line will select that peak. Double clicking on a peak list line will cause a view to center on that peak. And selecting a peak in a view will cause peak lists to scroll to that peak.

Resonance peak lists

You can display a peak list containing peaks from all spectra that are assigned with a given resonance. There are several ways to get this list. You can press the Resonance Peaks button in the assignment dialog (at). You can use the command r2 which bring up the list with for the w2 resonance of the currently selected peak. The commands r1 and r3 do the same for axes w1 and w3. If you click on a chemical shift in an assignment table (tb) the list for that resonance is shown. Or you can use the "resonance assignments" command ra and select a condition and a resonance using the entry fields and menus. (The menus pull down from the odd V character to the right of the entry field. Yeah it's weird.)

Peak list options

Crossdiagonal peaks

There is a peak list option called "Pair crossdiagonal peaks?" If you turn this on, peaks in 2-D homonuclear spectra that are assigned on both sides of the diagonal are placed on adjacent lines in the peak list. This facilitates comparing volumes or other properties.

Mardigras and Dyana formats

Sparky peak lists can be output to a file in MARDIGRAS or DYANA format. You select "Mardigras format" or "Dyana format" from the peak list options dialog. Turn off the display of all other fields unless you want these in the output file. The DYANA format puts the peak note in the last column which is intended to indicate "weak", "medium", or "strong". There will be blank lines in the list for unassigned peaks. These are place holders. Each peak in a spectrum always occupies a line in the peak list even if no data is displayed. Blank lines are filtered out when you write a peak list to a file.

Peak Notes

You can attach a note to a peak to record unusual features. You place a note on the selected peak(s) using the note dialog (nt). The Apply or Ok buttons set the note for all selected peaks. The notes can be displayed and searched for in peak lists.

Picking Peaks

Use the find/add peak pointer mode (F8) to pick peaks. Clicking on a point in a spectrum view will place a peak marker there. You can move the marker to the local maxima with the peak center command (pc). Dragging a box will place peak markers at local maxima satisfying minimum height, linewidth and drop off requirements. You can pick peaks over the whole spectrum using the region dialog (rt). The positive and negative height thresholds equal the lowest positive and negative contour level. You can adjust the contour levels with the contour scale (vC) or the contour dialog (ct). The minimum linewidth threshold is set with the peak picking dialog (kt). The default value is zero which has the effect of no minimum linewidth. The linewidth for a prospective peak is computed as the full width at half height. The minimum drop off requirement is a way of filtering out overlapped peaks. If the data height does not drop off by a specified amount as you move along each axis from the peak marker before starting to rise then that peak is not picked. The specified amount is given as a fraction of the total peak height. The default value is zero so there is no dropoff requirement by default.

Selecting Peaks

You can select peaks using the "select" pointer mode (F1). Select individual peaks by clicking on them with the left button. Drag a box around a group of peaks to select all of them. Selecting a peak or group of peaks causes all other peaks to be unselected. To avoid unselecting the currently selected peaks, hold the shift key while pressing the left mouse button. Pressing the shift key and clicking on a selected peak will deselect it without deselecting other selected peaks.

There are commands to select specific groups of peaks in the currently selected spectrum.

All peakspa
Fully assignedpF
Partially assignedpP
UnassignedpN
Select peaks by colorpC
Invert peak selectionpI
Aliased peakspA
Intra-residue assignmentspR
Sequential assignmentspS
Medium range assignmentspM
Long range assignmentspL

The select peaks by color command finds all peaks of the same color as a currently selected peak. Sequential assignments means i to i+1. Medium range assignments are i to i+2, i+3, or i+4. Long range assignments connect atoms 5 or more residues apart. All of the selection commands except select peaks by color unselects all previously selected peaks.

Grids and lines

You can draw noesy walks using lines. Select the "add line" pointer mode. Now click on a position to start a line and another position to end the line. Lines are forced to be vertical or horizontal. Grids are lines that extend from edge to edge of a spectrum and are intended to be used for looking at alignments. Select the pointer mode "add grid horz", "add grid vert", or "add grid both" and then click on a point to place a grid. Lines and grids appear in all views of the spectrum on which they are placed.

Ornament color, size, locking

"Ornaments" are peaks, labels, lines, grids, peakgroups, ie. spectrum annotations. You can change their color and size using the ornament dialog (ot). You can lock peaks (pk) so that they are not moved accidentally with the mouse, or during integration. Peak, label, line, and grid positions can also be locked with the ornament dialog. If you attempt to move a locked ornament, it will not move and a warning will be written in the main Sparky window.