No part of this document may be reproduced, transmitted, transcribed, stored in a retrieval system, or translated into any language in any form by any means without the written permission of Spectrum Research, LLC.

Spectrum Research, LLC. reserves the right to change the information in this document without prior notice.

Trademarks

SpecMan and NMR-SAMS are trademarks of Spectrum Research, LLC.

Acknowledgments

Computer-Assisted Structure Elucidation of Paclitaxel (Taxol) Using SpecMan and NMR-SAMS

Overview:

The following document provides step-by-step instructions that lead you through the process of Computer-assisted Structure Elucidation of the Paclitaxel (Taxol, Fig.1) molecule. This document is intended for the day-to-day users of SpecMan and NMR-SAMS and we assume that the users of this document have a good understanding of general NMR techniques and its application to structure elucidation. There are two parts to this document. Part I describes the step-by-step instructions that lead through each task from beginning to end while working with SpecMan. Part II describes the step-by-step instructions that lead through each task from beginning to end while working with NMR-SAMS.

Figure 1. Two-dimensional structure of Paclitaxel (taxol) with the ¹³C and ¹H (in parenthesis) resonance assignments. The numberings of the atoms corresponds to those of the ¹³C and ¹H peaks in Tables I and II. The three phenyl groups are not considered during the computer-assisted structure elucidation.

Part I: Computer-Assisted NMR Analysis and Assignment with SpecMan.

This part provides step-by-step instructions for computer-assisted peak picking with SpecMan.

I-1 Transferring Processed Spectrum from NMR Spectrometers

The sample was provided by Dr. R. Holton in the Chemistry Dept. of Florida State University. All of the 1D and 2D NMR data for this molecule were collected on Bruker DMX 300 MHz spectrometer. Solvent CDCl₃ was used in all experiments. Some of the spectra were obtained from different sample conditions, and therefore slight chemical shift differences are observed between the spectra. The processed files from Bruker are transferred via ftp or other means to the SGI workstation. In order to import Bruker processed spectra into SpecMan, one needs to transfer the 1r/2rr files along with their corresponding procs and proc2s files to the working directory on the SGI workstation. For this example, a main working directory called Paclitaxel was created with the following sub-directories:

H-1/

C-13/

DEPT-45/

DEPT-90/

DEPT-135/

HMQC/

DQFCOSY/

HMBC/

Next, the 1r/2rr and procs/proc2s files of each NMR experiment were transferred into the corresponding sub-directories from the sample data directory described below.

The default location for the sample data of SpecMan and NMR-SAMS is: /usr/share/Spectrum/Data/SpecMan/paclitaxel for SpecMan data, and /usr/share/Spectrum/Data/NMR-SAMS/paclitaxel for NMR-SAMS data. The NMR experimental data is located in the sub-directories (H-1 through NOESY). Before proceeding further please make a backup copy of the distribution data that was sent with the demo software.

I-2 Analysis of 1D Proton Spectrum

At the UNIX prompt, type specman to launch the SpecMan program. Next, Open the 1D file by selecting Open Spectrum option in the File menu. The dialog box that is shown below appears and select the File Type as Bruker, Then use the file browser to change directory to H-1 and double click on the 1r file in this directory. The 1D spectrum will be displayed in a 1D Window as seen below. The first step will be to set the reference. Although the reference parameters are obtained from the procs file, the user may want to change the spectral reference.

Setting Reference

Before setting reference, zoom the desired peak by using the right mouse button which activates a rubber-band zoom option. To zoom a peak, use the right mouse button to select the left top corner and drag the mouse (keeping the right mouse button pressed) to the desired right bottom corner (a rectangular zoom box will be drawn as you do this action) and release the button. The peak that you selected will be expanded and redrawn in the window. In this 1D spectrum you will set the reference on a small, weak peak due to CHCl₃at about 7PPM. To zoom this peak, keep the shift key on the keyboard pressed and then use the right while mouse button as described above to zoom this peak. You will notice the zoom being applied with fixed vertical scales. If the you want to further increase the vertical expansion of the peak keep the Ctrl key pressed while using the right mouse button for rubber-band zooming. This will help you adjust the vertical scales to the desired extent. After zooming, select the Set Reference option in the Edit Menu. Place the cross hair cursor at the top of the peak and click the left button. A Set 1D Reference dialog box appears with the X chemical shift of the current location. Type 7.27 in the X Reference PPM. Then click OK to accept the new reference. To reset the expansion to full view, select Reset Zoom (in the Display Menu) or the Reset Zoom icon in the tool bar. Once reference is set, the relevant spectral parameters will be saved with the data file.

In vast majority of cases peak picking of proton NMR spectrum is pointless, and without a very complicated analysis, one is not going to get what one needs, proton chemical shifts, from such a peak pick. For second order spectra the analysis becomes even more tedious. So it is not just a consequence of severe overlap in ¹H peaks which makes it difficult to do peak picking directly on this spectrum, but also other reasons as stated above. Hence, it is preferable to use 2D HMQC to extract 1D proton chemical shifts.

I-3 Analysis of 1D Carbon Spectrum.

Setting Reference and Appropriate Threshold

After opening the 1r file from the C-13 directory, set reference in a similar manner as described for 1D ¹H. The reference is set on the CHC1₃ peak at 77.22 ppm. Next do the 1D peak picking of ¹³C spectrum. Before peak picking, set the appropriate threshold. Select the Set Threshold button in the 1D Control Panel, which is displayed below. This will activate a red horizontal cursor on the 1D spectral window as the mouse is moved in this window. Move this horizontal cursor to a suitable position so that it is just above the noise level. Then click the left mouse button to update the threshold in the 1D Control Panel. One can also type the actual value in the threshold box. For this example use 2.381e+07 for threshold.

Automatic Peak Picking

Choose the Pick Peaks Automatically from the Analysis menu, then choose 1D from the Pick Peaks Automatically pull-right menu. This activates a Pick 1D Peak dialog box shown below. You can also activate the 1D peak picking by selecting the 1D peak picking icon in the tool bar.

Turn off the Negative Peak option and then click OK. After peak picking, the 40 picked peaks are displayed on the spectrum, and are also listed in the following 1D Peaks Table.

This peaks table can be resized, and the two sections can be made larger or smaller with the small hash bar in the middle of the table. The four main peaks table operations can be performed via the buttons at the bottom of the table. Please see the online help for more information about working with peaks tables.

Manually Removing and Adding Peaks

Next, delete the three strong solvent peaks by choosing the Remove Peaks option from the Analysis menu. After activating Remove Peaks the program prompts the user to define a box around the peaks that need to be deleted. A rectangular rubber-band zoom box is defined using the left mouse button (enclosing peaks 18-20) to remove the peaks. After removing the three solvent peaks (peak numbers 18-20), deactivate the Remove Peaks button in the Analysis menu by selecting it again. SpecMan also has an Undo option which allows restoring peaks that have been removed accidentally. The Undo option is activated by either selecting Undo in the Edit Menu or the second icon from the left on the tool bar.

The large peak at 167.25 PPM, is an envelope of two overlapping peaks. To see this better, first zoom into the dominant peak. Then, add a peak to this one by first selecting Add Peaks Manually from the Analysis menu and then choosing the Without Refine option from the Add Peaks Manually pull-right menu. After selecting this option, click the cursor at the place where the smaller peak can be seen affecting the shape of the dominant peak to add a peak. Next, deactivate Add Peaks Manually by selecting it again along with the pull-right menu option.

Sorting, Editing and Saving Peaks Table

Next select Edit Table button in the 1D peaks table, and a edit dialog box (shown below) appears. Complete the dialog box as follows: Select Sort table Entries button. Select Descending radio button and X Value radio button to sort the table in the descending order of X Values (i.e. ¹³C Chemical shifts). Select the Renumber Table ID's button on. Then click OK to sort the table. The 38 peaks in the table are sorted and renumbered in the descending order of their chemical shifts.

Next click Save Table in the 1D peaks table, and type a filename c13 to save the peaks in a table called c13.pks. This table will be saved in the same sub-directory. (The extension ".pks" is automatically added).

I-4 Analysis of DEPT Spectrum

In order to get the ¹³C multiplicity information, we need to use only two DEPT experiments: DEPT-90 and DEPT-135. In some instances, DEPT-45 may be used to detect potential errors such as missing peaks. For this example, we do the peak picking in all three spectra (DEPT-45, DEPT-90 and DEPT-135), but use only the DEPT-90 and DEPT-135 peak lists along with ¹³C peak list for the multiplicity analysis in NMR-SAMS.

Setting Reference and Appropriate Threshold

Similar to the ¹³C spectrum, the DEPT spectra are analyzed one at a time. First the DEPT-45 spectrum is opened and displayed for setting the reference. The second peak from the right in the spectrum is set to 15.0268 PPM according to the following steps:

Select Peaks Table with pull-right 1D option in the Analysis menu. A 1D peaks table will be displayed (this could be an empty one if the previous peaks table was cleared and closed). Next click Load Table in the 1D Peaks Table panel and select c13.pks file in the ¹³C sub-directory. The ¹³C peak positions will be annotated on the DEPT spectrum. This can be used as a guide to set reference on the DEPT peak. If the peak top of the second peak from the right in the DEPT-45 spectrum matches with the peak symbol (shown as “plus”) corresponding to ¹³C peak at 15.0268 PPM, then the reference is already set, and you can skip rest of this section to proceed with the peak picking step described in the next section. If the peak top is shifted from the ¹³C reference peak, then select Set Reference from the Display menu, and place the cursor on this peak. Next keeping the left mouse button pressed (a symbol in the form of a “plus” enclosed within a circle appears to mark the peak position selected for setting the reference), drag the cursor and release it at the center of the “plus” peak symbol which shows the position (which is the peak symbol at 15.0268 PPM). After releasing the left mouse button, a dialog box (similar to the one shown in Page 9) is displayed, showing the current chemical shift. Click OK and the reference of the DEPT-45 spectrum will be the same as that of the ¹³C spectrum.

Peak Picking, Removing Peaks and Manually Adding Peaks

Next perform peak picking of DEPT-45 spectrum. Set the appropriate threshold as described before (or type 8.183e+07 in the threshold box on 1D Control panel). Select auto 1D peak picking from the Analysis menu and the program picks 25 peaks. Next sort the peaks in the peaks table in the descending order of X Value as described previously. The sorted peaks are saved as dept45.pks by using the Save command in the 1D peaks table.

Next process the DEPT-90 and DEPT-135 spectra in a similar way. Remember to first set the reference in these spectra by following the same procedure as described above for DEPT-45. For DEPT-90 use the threshold 2.154e+07 and 15 peaks will be picked by auto 1D peak picking. For DEPT-135 use the threshold 3.417e+07 and also turn the Negative Peaks button on in the 1D peak picking dialog box (because there are negative peaks in DEPT-135). 23 peaks will be picked in the DEPT-135 spectrum. Make sure to manually add the shoulder peak at 35.84 ppm. As before Sort and Save the peaks as dept90.pks and dept135.pks, respectively.

I-5 Analysis of HMQC Spectrum.

HMQC spectrum provides both C-H direct connectivity information and the ¹H chemical shifts. Due to peak overlap in 1D proton spectrum, HMQC spectrum is used to extract 1D ¹H chemical shifts. This process is described in the next few sections.

Setting Threshold

The 2rr file of HMQC spectrum is opened by selecting Open Spectrum in the File menu. If the import is being performed for the first time, contours will be generated on the fly based on a computed threshold. The spectrum is displayed in the main SpecMan graphics window along with a Threshold palette (both displayed below).

The Threshold palette (shown above) is used to set appropriate threshold, number of contour levels and the contour level separation. These controls can be adjusted interactively with the sliders. Before changing the threshold, turn off the Auto Redraw button in the Threshold palette. Auto redraw can be turned on when working with smaller data sets which are sub-matrices of a large 2D spectra. Another way to set appropriate threshold is by stepping through the Starting Level slider. Stepping through levels to determine the threshold is useful when the spectrum has severe t₁ or t₂ noise ridges.

After adjusting the starting level, click Update to re-generate the contours with the new threshold. Also one can adjust the Separation and Number of Levels to get a more satisfactory display of the peaks. For this spectrum, use the Threshold as 4.749e+06, Separation as 1.2, and the Number of Levels as 20.

Setting Spectral Reference

Next set spectral reference by selecting the Associate Reference Spectrum in the Display menu. This option allows simultaneous display of 1D reference spectra on a 2D spectrum. The ¹H reference spectrum is displayed along the X axis (F2 dimension), and the ¹³C reference spectrum (or DEPT-45) is displayed along the Y-axis (F1 dimension). When this option is selected, a dialog box appears (as seen on the next page) and prompts for the file names of reference spectra and reference peaks table. The details of using this dialog box can be seen by selecting the Help button on this dialog box.

After selecting the appropriate reference spectra and its corresponding peaks table, click OK to display the reference spectra and the ¹³C grid lines. ¹H peaks table is not available now; otherwise it can also be displayed as grid lines. The grid lines are drawn at the coordinates of the 1D ¹³C peaks, which indicate the plausible locations of the 2D peaks in the HMQC spectrum. This will help the user verify the results of peak picking as seen below.

To set the reference of the HMQC, first zoom on a well-resolved cross peak with its corresponding 1D ¹H and ¹³C peaks displayed. Check if both the 1D ¹H and ¹³C peaks are aligned with the center of this cross peak. If the 1D’s are not properly aligned, you must set reference on 2D HMQC again to match these 1D peak locations. Select Set Reference in the Edit Menu and a cross hair cursor will appear. Move the cross hair cursor intersection to the center of the cross peak and press the left mouse button, and keeping it pressed (as soon as you press the left mouse button, the selected position is marked with a symbol that has a “plus” within a circle) drag it to the intersection point of the 1D ¹H and ¹³C peak coordinates in the 2D spectra so that it matches with the intersection, and release the left mouse button. The Set 2D Reference dialog box will appear (as shown below) with the new X and Y reference PPM for that location.

Click OK in the dialog box to accept these new reference values. This will set the reference on the selected cross peak.

Cross Checking 1D Peaks with 2D to Identify and Add Missing Peaks

Working with grid lines created from 1D chemical shifts has many advantages. It provides a nice way to verify 2D peak picking results, and also identifies missing 1D peaks if there are any by comparing with the 2D cross peaks. For example, in this spectrum, you will notice that there is no corresponding grid line (i.e. 1D ¹³C chemical shift) for the two HMQC peaks at about (H1: 4.25, C13: 76.78). By checking the DEPT-45, it can be seen that a ¹³C peak is hidden by a strong solvent peak of CHCl₃. Next add the ¹³C missing peak to the ¹³C peaks table as described below:

First load the 1D ¹³C spectrum by selecting Open Spectrum in the file menu. Next load the DEPT-45 peaks table by selecting the Load Table button in the 1D Peaks Table and choosing file dept45.pks in the sub-directory DEPT-45. The DEPT-45 peaks will be displayed on the 1D ¹³C spectrum. The chemical shift of the ¹³C peak buried under one of the solvent peaks is seen at 76.6756 ppm. Next load the ¹³C peaks table and select Add Peaks Manually with the pull-right option Without Refine in the Analysis menu. Move the cursor around the solvent peak until the chemical shift displayed on the status bar reads 76.6756ppm. At that point click the left mouse button to add a peak. Next sort the peak list in the descending order of ¹³C chemical shifts and save the new peak list in the same file c13.pks.

Correcting Chemical Shift Reference Offset between 1D and 2D

It is important to verify the alignment of 1D and 2D peak coordinates. Do the following steps to verify and correct the alignment as needed. First zoom on the cross peak which is at the lower most left corner of the 2D spectrum and check if both the ¹H and ¹³C peak are aligned well with the center of this cross peak. Set reference on that peak and then move the cross hair to a cross peak which is at the upper most right corner of the 2D spectrum. Check if both the ¹H and ¹³C peak are well aligned with the center of this cross peak. If there is a discrepancy, then the Sweep width along X or Y needs to be adjusted. This correction in Sweep Width can be applied by selecting the Spectral Parameters option in the display menu. The following spectral parameter dialog box appears, enter the new Sweep Widths along X or Y in this dialog box..

In the paclitaxel HMQC data the ¹H chemical shifts do not match well. There is a discrepancy of about 0.08 PPM between the 1D and 2D ¹H spectral width. After changing the Sweep Width along X from 7.9787 to 7.9700 PPM the peaks are well aligned in the 1D and 2D.

Note that the modified parameters are saved by SpecMan so the next time the spectrum is opened, such correction does not need to be repeated.

Auto Peak Picking of Cross Peak Multiplets.

Peak picking of HMQC data consists of auto peak picking followed by manual editing of the peak list. SpecMan uses a novel peak picking algorithm to automatically pick the center of mass of cross peak multiplets. To start peak picking select Pick Peak Automatically in the Analysis menu with the 2D option in the Pick Peak Automatically pull-right menu. A dialog box appears with various peak picking options as shown below.

Turn off the Negative Peaks button as well as the Grid Intelligence option. Choose the Merge Peak Multiplets option to pick center of mass of cross peak multiplets. The Merge Peak Multiplets option can be performed in three modes: Average, Weighted Average and Highest Peak. Select the Weighted Average in the Multiplet Picking Option.

SpecMan peak picking algorithm uses peak width filters to pick cross peaks as a merged multiplets with a center of mass. These filters are defined in terms of a minimum and maximum box size for the search algorithm. The minimum box size is used to filter noise peaks. The maximum box size corresponds to the width of an average cross peak multiplet. The limits can be determined by using the “Set Graphically” button in the Pick 2D Peaks dialog box. Upon selecting this option SpecMan will prompt the user to draw a rectangular box with the left mouse button on the peak which could be either a noise peak or a real cross peak for setting minimum and maximum limits respectively. For this data, enter 0.03 and 0.3 as Minimum X and Minimum Y for filtering noise peaks, and 0.08 and 1.0 as Maximum X and Maximum Y for merging multiplets. Next click OK to pick peaks. 26 peaks are automatically picked by SpecMan and the 2D Peak Table will be displayed as shown here.

Manual Editing of Peak Picking Results.

For efficient structure elucidation, it is always important to refine the peak picking results carefully. SpecMan provides a nice peak table interaction feature for the user to examine the zoomed 2D peaks one at a time. To do this the user first zooms on a particular peak of interest and then selects the peak entry in the 2D peaks table to browse through the next peak. In this manner the user steps through every peak entry in the peaks table and SpecMan automatically zooms and displays the selected peak. While examining a zoomed peaks in the 2D spectrum, the user can compare its center with the grid lines which indicate the expected positions of the peak centers.

For this example, some odd cross peaks which exceed the maximum box size are missed and these are added manually by choosing Add Peaks Manually from the Analysis menu with the pull-right option Without Refine. After activating Add Peaks Manually, move the pointer cursor to the center of the cross peak and click the left mouse button to pick the peak. The peak positions of the two partially overlapping peaks at about (¹H: 2.33, ¹³C: 35.94) are corrected in this manner. Several other peaks are also corrected. The ¹H coordinates of the HMQC peaks will be used as the basis for peak picking in other spectra, and therefore it is important to pick them as accurately as possible. The down-field cross peaks that arise from the phenyl groups are not carefully cleaned here because they are going to be ignored by NMR-SAMS during the partial structure elucidation. Finally 28 HMBC peaks are obtained.

Sorting and Saving Peaks, Extracting ¹H Chemical Shifts

Click Edit Table button in the 2D Peaks Table and a Edit Peaks Table dialog box appears as shown here.

Choose Sort Table Entries according to Descending, Y Value (i.e., ¹³C chemical shift) as the Sort Key and select Renumber Table ID's to renumber after the sort. Also, select Extract coordinates to 1D Table along the X dimension. After selecting the above options, click OK. SpecMan sorts the 28 HMQC and creates a 1D peaks table of 28 ¹H chemical shifts. Sort the ¹H peaks in the 1D peaks table by selecting Sort Table Entries according to Descending, X Value (i.e., ¹H chemical shift) as the Sort Key and select Renumber Table ID's to renumber after the sort. Next select the Save button in the 1D peaks table and use the file name h1 to save the 1D H¹ peaks. A file with h1.pks name will be saved in the current directory. Also, save the 2D HMQC peaks table as hmqc.pks.

Adding ¹H Peaks of Heteroatom-attached Protons

The ¹H peaks table extracted from the HMQC peaks table does not include the protons attached to heteroatoms. To make the ¹H peaks table more complete, open and display the ¹H spectrum as described previously, and load the ¹H peaks table by selecting Load Table button in the 1D peaks table. (Choose Peaks Tables with pull-right option 1D in the Analysis menu if the 1D Peaks Table is not opened). The picked ¹H peaks are now annotated on the ¹H spectrum. The unlabeled peaks at 7.0, 3.58, 2.45, and 1.74 ppm respectively, are added by choosing Add Peaks Manually with the pull-right option Without Refine in the Analysis menu. The expanded peaks table, which now contains 32 peaks, is sorted and saved as h1.pks again.

I-6 Analysis of DQF-COSY Spectrum.

Setting Spectral Reference and Threshold

Using procedures similar to the ones described under analysis of HMQC, the DQF-COSY spectrum is opened, displayed with appropriate threshold and spectral reference. The threshold for DQF-COSY is set at 2.730e+06, with contour Separation as 1.3, and Number of Levels as 20.

It is recommended to use a fairly low threshold of DQF-COSY so that all the weak COSY peaks appear. This will increase the efficiency of structure generation in NMR-SAMS. The reason is that NMR-SAMS uses negative information from DQF-COSY which is an additional constraint for structure elucidation. For example, two proton-bearing carbons would be forbidden to connect if their protons show no DQF-COSY peaks. To take advantage of this feature, one has to make sure that all the COSY peaks, including the very weak ones, are picked.

Grid Intelligence-based Peak Picking

As described before in the HMQC analysis, select Associate Reference Spectrum from the Display menu to display the 1D ¹H reference spectrum along both X and Y axis. Also select the ¹H chemical shifts table (h1.pks) to display the grid lines on the DQF-COSY spectrum. The spectral reference of the DQF-COSY spectrum is calibrated in the same manner as the HMQC spectrum by checking the alignment of the 1D peaks with the 2D peaks. The spectral width of F1 dimension is adjusted from 8.2636 to 8.2550 to get a better match between the 1D and F1 ¹H chemical shifts.

The Peaks of DQF-COSY were picked with a novel peak picking method called “grid intelligence-based peak picking.” In this procedure the grid intersection points, instead of the minimum peak width, are used as filter in the search algorithm for peak picking. After a multiplet has been merged and the center of mass is calculated using the current peak merge option, SpecMan attempts to locate a grid intersection point within the specified minimum peak width box size. If such a grid point is found, the peak center is retained as a real peak. Otherwise, it is rejected. To use this method for paclitaxel, do the following steps:

Select Pick Peaks Automatically with 2D pull right option in the Analysis menu. This will bring up a Pick 2D Peaks dialog box as shown below. Keep the Negative Peaks option on (because for COSY you should include the negative peaks too). The Ignore Diagonal Peak option is not used as the diagonal peaks will be automatically removed by NMR-SAMS in the subsequent analysis. Select Grid Intelligence option and set 0.08 for both Minimum X PPM and Y PPM under Grid Distance Filter. You can also set these limits graphically by selecting the “Set Graphically” button in the Pick 2D Peaks dialog box. Next select the Merge Peak Multiplets option and set the mode as Weighted Average to process multiplets. Under the Peak Width Filter, set 0.001 for Minimum X PPM and Minimum Y PPM as limits for discriminating noise peaks, and set 0.15 for both Maximum X PPM and Maximum Y PPM as the maximum peak width for considering a cross peak multiplet. Click OK, and 49 peaks will be picked less than 1 minutes. For more details on using the Use Grid Intelligence method refer to the on-line help of SpecMan.

After the peak picking, it is important to examine the picked peaks and correct the ones which are either missed or not picked properly. The display of the reference 1D spectra as well as the grid lines facilitates the user’s verification of the peak picking results to a great extent. Note that in the subsequent analysis steps, NMR-SAMS will automatically discard the diagonal peaks and merge symmetry-related peaks for a homonuclear spectrum. So it is not necessary to discard the diagonal peaks at this stage. Although it is not very important to pick the symmetric peaks (on either side of the diagonal), picking both the symmetric peaks may improve the reliability of the subsequent analysis. SpecMan picks both the symmetric peaks, and NMR-SAMS filters them appropriately during the conversion from SpecMan peak list to NMR-SAMS NMR data file. In cases where multiple grid centers are close to a peak and it is difficult to resolve the ambiguity, it is best advised to leave the picked peak as is, because NMR-SAMS will automatically include the different possible correlations to 1D peaks, and treat this cross peak as an ambiguous correlation. In this spectrum, some noise peaks are removed from the peak list. For this spectral data set, the cross peak at about (2.48, 4.41) is shifted from the 1D ¹H peak position (because of sample conditions) so it is not automatically picked. A peak at the grid intersection (2.47, 4.40) was manually added. Similarly the other peaks are examined and added, removed or corrected if necessary. After the cleaning, the peaks table has 50 peaks (includes some diagonal peaks). The peaks table is sorted and saved as a file "cosy.pks". If you have trouble cleaning the DQFCOSY data, you can load the cosy.pks peak list that was supplied with the sample data into SpecMan to see which peaks were retained. To do this, click Load Table… in the 2D Peak Table palette and select cosy.pks in the file browser. Then click OK and the clean COSY peaks will be annotated on the spectrum.

Note that, in addition to the parameters listed in the Pick 2D Peaks dialog box, the peak picking results also depend on the chemical shift reference and the alignment between the 1D and 2D spectra (when grid intelligence is used). If some of these are changed, you may get slightly different results, but it should not drastically affect the subsequent structure elucidation.

I-7 Analysis of HMBC Spectrum

Setting Spectral Reference and Threshold

Using procedures similar to the ones described under analysis of HMQC, the HMBC spectrum is opened, displayed with appropriate threshold and spectral reference. The threshold for HMBC is set at 7.639e+06, with contour Separation as 1.2, and Number of Levels as 20.

As described above, select Associate Reference Spectrum from the Display menu to display the 1D ¹H and ¹³C reference spectra along the X and Y axis, respectively. The ¹H and ¹³C peaks tables (h1.pks and c13.pks, respectively) are also displayed as grid lines on the HMBC spectrum. The spectral reference is calibrated in the same manner as the HMQC spectrum by checking the alignment of the 1D peaks with the 2D peaks. The spectral width of both dimensions are adjusted to get a better match between the 1D and 2D ¹H and ¹³C chemical shifts.

Grid Intelligence-based Peak Picking

Similar to that of DQF-COSY, the peaks of HMBC were picked with the grid intelligence- based method. Select Pick Peaks Automatically with 2D pull right option in the Analysis menu. This will bring up a Pick 2D Peaks dialog box. Keep the Negative Peaks option off (because for HMBC you need to include only the positive peaks) and select the Grid Intelligence button. Under Grid Distance Filter set 0.02 and 0.5 as Minimum X PPM and Y PPM, respectively. Next select the Merge Peak Multiplets option and set the mode as Weighted Average to process multiplets. Under Peak Width Filter, set 0.001 for Minimum X PPM, 0.02 for Y PPM as limits for discriminating noise peaks, and set 0.1 for Maximum X PPM and 2.0 for Y PPM. The maximum limits are used as the maximum peak width for considering a cross peak multiplet. Click OK, and 98 peaks will be picked in less than 1 minute. For more details on using the Use Grid Intelligence method refer to the on-line help of SpecMan.

I-8 Editing Peak Tables before Using NMR-SAMS

In both the ¹H and ¹³C spectra, some of the resonances due to aromatic atoms are very difficult to resolve. However, the three phenyl groups can be easily identified from other spectral data (e.g. IR, UV). In situations like this one can ignore the three phenyl groups and let NMR-SAMS use only the well-resolved spectral data for the core structure. This is called “partial structure elucidation” and will be demonstrated in this example.

Part II: Computer-Assisted Structure Elucidation by NMR-SAMS

II-1 Introduction

In Part I, using SpecMan, the peaks table for 1D ¹H, ¹³C, DEPT-45, DEPT-90, DEPT-135, and 2D DQF-COSY, HMQC, and HMBC spectra were obtained. In this part, the SpecMan peak lists are converted into correlation information between the 1D peaks, and then interpreted to define bond constraints on the atoms labeled by the 1D chemical shifts. These bond constrains are further cross-checked with the molecular formula to create a consistent set of C-C bond constraints. Based on the available data, a set of structural building blocks are generated, and an atom-atom connection matrix is setup. Finally the plausible structures are generated.

II-2 Getting Started with NMR-SAMS

At the UNIX prompt, type nmrsams to launch the NMR-SAMS program. The program starts with a Main Graphics Window that has a menu bar and another window called Status Window which displays text prompts to indicate the current status of the structure elucidation. The messages in the status window also tells the user the next possible steps to do with NMR-SAMS. In most of the steps during the structure elucidation, NMR-SAMS displays messages which are of four types: error, warning, information and interrogation corresponding to different situations. For this part of the tutorial use the SpecMan generated peak lists provided under the sample data directory of NMR-SAMS which is located in: /usr/share/Spectrum/Data/NMR-SAMS/paclitaxel.

II-3 Opening New Working Data Set

By selecting New in the File menu, a file browser is brought up (as shown below), listing all the files with extension of ".mdf" in the current sub-directory.

Enter a new root name for the working data set (say paclitaxel-test), click OK, and NMR-SAMS will create the following files in the current directory:

An empty master data file (with file extension .mdf), paclitaxel-test.mdf, where all the intermediate and final results are stored.

A default parameter file, called paclitaxel-test.par, where the control parameters used for the data interpretation and structure generation are stored. You can access the parameters by using the commands in the pull-right menu of Edit/Parameters.

An empty NMR data file, called paclitaxel-test.nmr, where the NMR data converted from the SpecMan peaks table will be stored. The user can edit this file by choosing NMR Data File in the Edit menu.

An empty log file, called paclitaxel-test.log, where most of the information, warning, and error messages produced during the analysis will be stored. The user can view the log file by choosing Log File in the Edit menu.

An empty structure file, called paclitaxel-test.str, where the connection table of the generated structures and their resonance assignments will be stored. The user can only view this file and delete or edit the contents. To view the file choose the Generated Structures option in the Display menu.

Next the user is prompted to enter the molecular formula. For this example enter C47H51NO14.

After clicking OK, the molecular formula is automatically interpreted for element composition and common valences. These are written into the MDF after the keyword “ATOMS:”.

Note: Suppose you do not know the molecular formula, you can type “unknown”. For structure elucidation with unknown molecular formula, please see section II-16.

II-4 Conversion of SpecMan ¹H Peak List.

First select Create NMR Data File with the pull-right option ¹H in the File menu, NMR-SAMS will open a dialog box (shown below) and prompt the user to enter the filename of the ¹H peaks table from SpecMan (In this case, select the h1.pks file).

Click OK to accept this dialog box, and you will be informed that 32 ¹H peaks are read.

Next the user is prompted (as shown in the information box below) to supply the ¹H multiplicity information if any. By selecting NMR Data File of the Edit menu, the multiplicity’s of 8 singlet peaks are defined as "s", 7 doublets as "d", and the remaining ones are defined either as "m" (for general multiplet) or “u” (for unknown). Table I shows these changes. Click OK and proceed with the next step.

Note that NMR-SAMS recognizes only the multiplet patterns such as singlet, double, triplet, and quartet. All others must be entered as unknown or multiplet in general. The multiplet information will be used to eliminate inappropriate bonds during structure generation.

Table I. The ¹H peak list of paclitaxel

H1: /usr/people/peng/NMR-SAMS/ndat/taxol-grid/h1-noar.pks

#1. 8.139 u ;1

#2. 7.747 u ;2

#3. 7.590 u ;3

#4. 7.551 u ;4

#5. 7.485 u ;5

#6. 7.478 u ;6

#7. 7.445 u ;7

#8. 7.363 u ;8

#9. 7.001 d ;9

#10. 6.285 s ;10

#11. 6.240 m ;11

#12. 5.796 m ;12

#13. 5.682 d ;13

#14. 4.946 m ;14

#15. 4.793 m ;15

#16. 4.403 m ;16

#17. 4.302 d ;17

#18. 4.207 d ;18

#19. 3.812 d ;19

#20. 3.584 d ;20

#21. 2.547 m ;21

#22. 2.453 d ;22

#23. 2.391 s ;23

#24. 2.361 m ;24

#25. 2.298 m ;25

#26. 2.236 s ;26

#27. 1.887 m ;27

#28. 1.801 s ;28

#29. 1.737 s ;29

#30. 1.695 s ;30

#31. 1.248 s ;31

#32. 1.152 s ;32

The first line which begins with the keyword “H1:” indicates the start of ¹H peak list. After the keyword “H1:”, following a blank space comments may be added up to 80 characters in length. The entries in the rest of the lines represent the Peak ID, chemical shift, multiplicity, and comments (optional) for each ¹H peak, respectively. One or more space(s) is used as a delimiter for all items except comments which are separated by “;”. Items marked as optional can be omitted unless an item following it is included. In such a case, please include default values for ignored items even if they don’t get used Comments can always be included as long as they follow a “;”. In this table, the number in the comment field corresponds to the ID of peak in the SpecMan peaks table.

II-5 Conversion of SpecMan ¹³C and DEPT Peak List.

Upon selecting Create NMR Data File with pull right option ¹³C and DEPT in the File menu, the following dialog box is displayed. Click Browse to enter the SpecMan C-13 Peaks Table, and select c13.pks in the file browser. Click the button named as DEPT, to select the Peak Multiplicity Experiment. Turn off the button named DEPT-45 Peaks Table, because it only provides redundant information, and we don’t need to use it when the other the other two experiments are being used. Click Browse to enter the DEPT-90 Peaks Table, and select dept90.pks in the file browser. Click Browse to enter the DEPT-135 Peaks Table, and select dept135.pks in the file browser. Enter 0.2 PPM for the matching tolerance between ¹³C and DEPT peaks.

Click OK and the following message box warns of an inconsistency detected for ¹³C peak #15. The is due to the unresolved aromatic peaks which are going to be ignored during the subsequent analysis. Click OK to ignore it, and the program will treat the multiplicity of that peak as “unknown”.

Next an information dialog box (as seen below) displays that 39 ¹³C peaks, with multiplicities, are obtained (see Table II). Click OK and proceed with the next step.

By comparison with the molecular formula, the program warns the user that there are fewer ¹³C peaks than expected (because the molecular formula has 47 carbons). As this example is for partial structure elucidation based on the well-resolved portion of the spectral data, click OK to ignore the warning message shown below.

Also it warns of the ¹³C peaks with unknown multiplicity. Click OK to ignore it.

If neither DEPT nor APT spectral data is available, the user can type only the ¹³C peaks table and select the None button for the peak multiplicity experiment. This will set the multiplicity as unknown for all the ¹³C peaks. One can also include the ¹³C peak multiplicity by manually editing the NMR data file. Without multiplicity information NMR-SAMS may fail to generate good structures. For more details see the on-line Help of NMR-SAMS.

Table II. The ¹³C peak list of paclitaxel

C13: /usr/people/peng/NMR-SAMS/ndat/taxol-grid/c13-noar.pks

#1. 203.826 s ;1

#2. 172.921 s ;2

#3. 171.427 s ;3

#4. 170.595 s ;4

#5. 167.252 s ;5

#6. 167.237 s ;6

#7. 142.202 s ;7

#8. 138.258 s ;8

#9. 133.914 d ;9

#10. 133.467 s ;10

#11. 132.157 d ;11

#12. 130.432 d ;12

#13. 129.415 s ;13

#14. 129.246 d ;14

#15. 128.938 u ;15

#16. 128.907 q ;16

#17. 128.568 d ;17

#18. 127.274 d ;18

#19. 84.646 d ;19

#20. 81.410 s ;20

#21. 79.331 s ;21

#22. 76.681 t ;22

#23. 75.803 d ;23

#24. 75.233 d ;24

#25. 73.477 d ;25

#26. 72.583 d ;26

#27. 72.398 d ;27

#28. 58.872 s ;28

#29. 55.282 d ;29

#30. 45.900 d ;30

#31. 43.404 s ;31

#32. 35.948 t ;32

#33. 35.855 t ;33

#34. 27.105 q ;34

#35. 22.837 q ;35

#36. 22.021 q ;36

#37. 21.035 q ;37

#38. 15.027 q ;38

#39. 9.773 q ;39

The first line which begins with the keyword “C13:” indicates the start of ¹³Cpeak list. After the keyword “ C13:”, following a blank space comments may be added up to 80 characters in length. The entries in the rest of the lines represent the Peak ID, chemical shift, multiplicity, and comments (optional) for each ¹³C peak, respectively. One or more space(s) is used as a delimiter for all items except comments which are separated by “;” Items marked as optional can be omitted unless an item following it is included. In such a case, please include default values for ignored items even if they don’t get used Comments can always be included as long as they follow a “;”. In this table, the number in the comment field corresponds to the ID of peak in the SpecMan peaks table.

II-6 Conversion of SpecMan COSY Peak List.

Upon selecting Create NMR Data File with pull right option COSY in the File menu, the user is prompted to enter the filenames of the COSY peaks table from SpecMan (cosy.pks), as well as a matching tolerance between 1D and 2D coordinates for each dimension as shown in the dialog box below. For this example 0.005 PPM is used for both dimensions.

After clicking OK, the user is informed (as shown in the information box below) that 17 COSY correlations are obtained. Also the user is prompted to denote some of the potential long-range coupled peaks as "weak" ones, while the program assumes all COSY as short range (geminal and vicinal) coupling by default.

In this example three DQF-COSY peaks are identified as potential long-range coupling. Peak #5 and #17 are identified as potential long-range coupling from the singlet ¹H peaks with which they interact. Peak #15 is identified as a potential long-range coupling due to its very weak peak intensity. Select NMR Data File in the Edit menu (This will put you in editor mode) to edit the NMR data file. Mark the three peaks as “weak” ones by modifying their intensity level to 1. (See Table III).

(Note: a potential long-range coupled COSY peaks is usually interpreted as 3-5 intervening bonds between the correlated protons, which also covers the possibility of vicinal coupling. A short-range coupling is usually interpreted as 2-3 intervening bonds between the correlated protons. If a long-range coupling is mistakenly interpreted as a short-range one, NMR-SAMS will not generate the correct structure. So it is very important to denote peaks due to potential long-range coupling. Geminal couplings are always automatically detected by the program.)

Table III. The COSY peak list of paclitaxel

COSY:

#1. (5 - 6) 3 0.00 ;3+5

#2. (9 - 12) 3 0.00 ;8+14

#3. (11 - 24) 3 0.00 ;10

#4. (11 - 25) 3 0.00 ;11

#5. (11 - 28) 1 0.00 ;12+46

#6. (12 - 15) 3 0.00 ;16+23

#7. (13 - 19) 3 0.00 ;18+33

#8. (14 - 21) 3 0.00 ;20+39

#9. (14 - 27) 3 0.00 ;21+43

#10. (15 - 20) 3 0.00 ;24+35

#11. (16 - 21) 3 0.00 ;27+37

#12. (16 - 22) 3 0.00 ;25+41

#13. (16 - 27) 3 0.00 ;28+42

#14. (17 - 18) 3 0.00 ;30+31

#15. (17 - 19) 1 0.00 ;29

#16. (21 - 27) 3 0.00 ;40+44

#17. (31 - 32) 1 0.00 ;47+49

The first line which begins with the keyword “COSY:” indicates the start of COSYpeak list. After the keyword “COSY:”, following a blank space comments may be added up to 80 characters in length. The entries in the rest of the lines represent the Cross Peak ID, IDs of the correlated 1D ¹H peaks shown in parenthesis (for ambiguous correlations the IDs of all possible 1D ¹H peak correlations are included), peak intensity levels (which are classified as four types: strong, medium, weak, and unknown, and denoted as 3, 2, 1, and 0 respectively. The default value is always 3.), J-coupling (optional, the default value is 0), and comments (optional, with a maximum size of 80 characters), for each COSY cross peak, respectively. For a short range coupled DQF-COSY peaks intensity levels should be either 3 or 2. For long-range coupled DQFCOSY peaks, the intensity levels should be 1. If an intensity level 0 used, the program will expect actual J-coupling values in the field which represents J-coupling. One or more space(s) is used as a delimiter for all items except comments which are separated by “;”. Items marked as optional can be omitted unless an item following it is included. In such a case, please include default values for ignored items even if they don’t get used. Comments can always be included as long as they follow a “;”. In this table, the numbers in the comment field correspond to the IDs of the peaks in the SpecMan peaks table. For merged peaks these numbers are shown with a + sign. For COSY peak list the comments are not currently used by NMR-SAMS.

II-7 Conversion of SpecMan HMQC Peak List.

Upon selecting Create NMR Data File with the pull-right option HMQC (or HETCOR) in the File menu, the user is prompted to enter the filenames of the HMQC peaks table from SpecMan (hmqc.pks) as shown in the dialog box below. Also the user is prompted to input a matching tolerance between 1D and 2D coordinates for each dimension. For this example, 0.01 PPM is used as tolerance for matching ¹H peaks and 0.2 PPM is used as tolerance for matching ¹³C peaks. The former is much smaller because the ¹H peak list was generated from the 1D projection of the HMQC peaks table.

The following message dialog box is displayed because the peak picking results in the aromatic region were not well-cleaned. Click OK to All to ignore all similar messages.

Upon completing the peaks table conversion, NMR-SAMS checks the results against the molecular formula, and shows the following warning message. Click OK to ignore it.

Also it warns about the ¹H peaks for which HMQC cross peaks were not observed. Click OK.

A peak list of 27 HMQC peaks are obtained and stored in the NMR data file. (see Table IV).

(Note: Unlike other 2D peaks, where ambiguous correlations are allowed, the HMQC peaks must have exactly two correlated 1D peaks).

Table IV. The HMQC peak list of paclitaxel

HMQC:

#1. (9 - 3) ;1

#2. (11 - 5) ;2

#3. (12 - 1) ;3

#4. (15 - 4) ;5

#5. (15 - 8) ;6

#6. (16 - 7) ;4

#7. (18 - 6) ;7

#8. (19 - 14) ;9

#9. (22 - 17) ;11

#10. (22 - 18) ;10

#11. (23 - 10) ;12

#12. (24 - 13) ;13

#13. (25 - 15) ;14

#14. (26 - 11) ;15

#15. (27 - 16) ;16

#16. (29 - 12) ;17

#17. (30 - 19) ;18

#18. (32 - 24) ;20

#19. (32 - 25) ;19

#20. (33 - 21) ;21

#21. (33 - 27) ;22

#22. (34 - 31) ;23

#23. (35 - 23) ;24

#24. (36 - 32) ;25

#25. (37 - 26) ;26

#26. (38 - 28) ;27

#27. (39 - 30) ;28

The first line which begins with the keyword “HMQC:” indicates the start of HMQCpeak list. After the keyword “HMQC:”, following a blank space comments may be added up to 80 characters in length. The entries in the rest of the lines represent the Cross Peak ID, IDs of the correlated 1D ¹³C & ¹H peaks shown in parenthesis, and comments (optional, with a maximum size of 80 characters), for each HMQC cross peak, respectively. One or more space(s) is used as a delimiter for all items except comments which are separated by “;”.

II-8 Conversion of SpecMan HMBC Peak List.

Upon selecting Create NMR Data File/HMBC (or COLOC) in the File menu, the user is prompted to enter the filenames of the HMBC peaks table from SpecMan (hmbc.pks) as shown below in the dialog box. Also the user is prompted to input a matching tolerance between 1D and 2D coordinates for each dimension. Here 0.005 PPM and 0.08PPM are used as tolerance for ¹H and ¹³C peaks respectively.

During the conversion, the user is warned (as shown below in the dialog box) if a cross peak picked with SpecMan is discarded because its ¹H or ¹³C chemical shift does not match a 1D ¹H or ¹³C peak, respectively, within the specified tolerance. For this example, click OK to All to accept them.

During the conversion process, the user is also warned of ambiguous peaks, which has more than one correlated 1D peaks within the matching tolerance in a dimension. Note that although NMR-SAMS can use ambiguous correlation information, too many ambiguous correlations will undermine the efficiency of the subsequent structure generation. So whenever possible the user should resolve ambiguities manually. For this example, click OK to All to accept all the warning messages. (In the dialog box below, the converted cross peaks is represented in a similar format as explained in Table V, except that the ¹³C and ¹H chemical shifts are also listed in the parentheses following the IDs of the correlated 1D peaks.)

After the conversion, 95 HMBC peaks (Table V) are obtained and saved in the NMR data file after a keyword “HMBC”..

Table V. The HMBC peak list of paclitaxel

HMBC:

#1. (1 - 10) 3 ;2

#2. (1 - 19) 3 ;1

#3. (1 - 30) 3 ;3

#4. (2 - 12) 3 ;6

#5. (2 - 15) 3 ;7

#6. (2 - 20) 3 ;5

#7. (3 - 10) 3 ;8

#8. (3 - 26) 3 ;9

#9. (4 - 23) 3 ;10

#10. (5 6 - 1) 3 ;13

#11. (5 6 - 2) 3 ;15

#12. (5 6 - 9) 3 ;11

#13. (5 6 - 12) 3 ;12

#14. (5 6 - 13) 3 ;14

#15. (7 - 10) 3 ;16

#16. (7 - 11) 3 ;17

#17. (7 - 24) 3 ;20

#18. (7 - 25) 3 ;19

#19. (7 - 28) 3 ;18

#20. (8 - 12) 3 ;21

#21. (8 - 15) 3 ;22

#22. (9 - 1) 3 ;23

#23. (10 - 5) 3 ;27

#24. (10 - 10) 3 ;24

#25. (10 - 11) 3 ;28

#26. (10 - 28) 3 ;29

#27. (10 - 31) 3 ;26

#28. (10 - 32) 3 ;25

#29. (11 - 2) 3 ;30

#30. (13 - 4) 3 ;31

#31. (15 16 - 8) 3 ;33+36

#32. (15 16 - 31) 3 ;32+37

#33. (15 16 - 32) 3 ;34+35

#34. (17 - 8) 3 ;38

#35. (18 - 2) 3 ;40

#36. (18 - 5) 3 ;41

#37. (18 - 12) 3 ;39

#38. (19 - 17) 3 ;46

#39. (19 - 18) 3 ;42

#40. (19 - 21) 3 ;44

#41. (19 - 22) 3 ;43

#42. (19 - 27) 3 ;45

#43. (20 - 14) 3 ;49

#44. (20 - 17) 3 ;48

#45. (20 - 18) 3 ;52

#46. (20 - 19) 3 ;47

#47. (20 - 21) 3 ;50

#48. (20 - 27) 3 ;51

#49. (21 - 13) 3 ;57

#50. (21 - 19) 3 ;59

#51. (21 - 24) 3 ;55

#52. (21 - 25) 3 ;53

#53. (21 - 29) 3 ;54

#54. (21 - 31) 3 ;58

#55. (21 - 32) 3 ;56

#56. (22 - 19) 3 ;60

#57. (24 - 19) 3 ;61

#58. (24 - 24) 3 ;64

#59. (24 - 25) 3 ;62

#60. (24 - 29) 3 ;63

#61. (25 - 12) 3 ;66

#62. (26 - 25) 3 ;67

#63. (26 - 28) 3 ;68

#64. (27 - 19) 3 ;73

#65. (27 - 21) 3 ;72

#66. (27 - 22) 3 ;69

#67. (27 - 27) 3 ;71

#68. (27 - 30) 3 ;70

#69. (28 - 13) 3 ;74

#70. (28 - 19) 3 ;75

#71. (28 - 21) 3 ;77

#72. (28 - 30) 3 ;76

#73. (29 - 5) 3 ;80

#74. (29 - 6) 3 ;82

#75. (29 - 9) 3 ;78

#76. (29 - 15) 3 ;79

#77. (29 - 20) 3 ;81

#78. (30 - 13) 3 ;83

#79. (30 - 16) 3 ;86

#80. (30 - 17) 3 ;84

#81. (30 - 18) 3 ;85

#82. (30 - 30) 3 ;87

#83. (31 - 10) 3 ;91

#84. (31 - 24) 3 ;89

#85. (31 - 29) 3 ;88

#86. (31 - 31) 3 ;92

#87. (31 - 32) 3 ;90

#88. (32 33 - 13) 3 ;93

#89. (32 33 - 22) 3 ;94

#90. (32 33 - 29) 3 ;95

#91. (33 - 16) 3 ;96

#92. (34 - 32) 3 ;97

#93. (36 - 31) 3 ;98

#94. (39 - 16) 3 ;100

#95. (39 - 19) 3 ;99

The first line which begins with the keyword “HMBC:” indicates the start of HMBCpeak list. After the keyword “HMBC:”, following a blank space comments may be added up to 80 characters in length. The entries in the rest of the lines represent the Cross Peak ID, IDs of the correlated 1D ¹³C & ¹H peaks shown in parenthesis (for ambiguous correlations the IDs of all possible 1D ¹³C & ¹H peaks are included), peak intensity levels (which are classified as four types: strong, medium, weak, and unknown, and denoted as 3, 2, 1, and 0 respectively. The default value is always 3.), and comments (optional, with a maximum size of 80 characters), for each HMBC cross peak, respectively. One or more space(s) is used as a delimiter for all items except comments which are separated by “;”. Items marked as optional can be omitted.

II-9 Generation of Building Blocks

Select the Building Blocks option in the Analysis menu to perform this task. In this step, the MF, ¹H, ¹³C, and HMQC spectral data are interpreted as possible set(s) of structural building blocks for structure generation.

MF is interpreted as the elemental composition of the unknown molecule. The results are written in to the .mdf file under the keyword “MF:”.

NMR-SAMS uses the peak lists in the NMR data file (.nmr file) and saves the results in the master data file (.mdf file). In the following sections the results of NMR data interpretation are described for each type of peak list. All of these actions happen after activating the Bond Constraints option in the Analysis menu.

¹H Peaks:

The chemical shifts and the multiplicity’s are listed in the MDF after a keyword "1DH1". The integral information, if any, is not used by the program now.

¹³C Peaks:

The chemical shifts and number of attached protons (derived from the ¹³C multiplicity) are listed in the MDF after a keyword "1DC13". The number of ¹³C peaks is compared with the number of constituent carbon atoms to determine the symmetry of the molecule. In this case fewer ¹³C peaks than the constituent carbon atoms are used, so the user is alerted (as shown below in the dialog box) that partial structure elucidation will be performed for paclitaxel. That is to say, NMR-SAMS will only consider the carbon atoms labeled by the available ¹³C peaks and generate partial structures (or substructures) based on them. Click OK to accept it.

HMQC Peaks:

Each peak is interpreted as one-bond connectivity between the relevant ¹³C and ¹H spins. In this case 20 bond constraints are listed in the MDF after a keyword "HMQC".

If a ¹H peak is found to have no HMQC peak, the user will be prompted to supply the type of heteroatom attached to it. The program then assigns a heteroatom to the proton and adds a bond constraint together with the HMQC ones. If the user chooses not to add such connectivity between the ¹H peaks and heteroatoms, all other connectivity information (e.g. COSY and HMBC) concerning these ¹H peaks will be ignored and this could result in increased computation time and plausible candidate structures.

For this example, first a dialog box (as shown below) informs the user that 5 ¹H peak do not show C-H connectivity in the HMQC spectrum, and lists the default assignment of these ¹H peaks to the composite heteroatoms.

Since the ¹H peak #2 (7.747ppm) is due to an aromatic proton, we can not accept this result. Click No to assign the ¹H peaks one by one. In the upcoming dialog boxes, type “unknown” for the first proton (#2), and accept the recommended values for the remaining protons (#9, #20, #22, and #29).

Upon completing this step, the building blocks are displayed in the main graphics window as show below. The #15 building block, represented as “CH?”, has uncertain number of attached protons because its ¹³C multiplicity is unknown. Building blocks #40 - #47 are also represented as “CH?” because no ¹³C peaks were observed for them (because of peaks overlap). These building blocks actually correspond to the aromatic carbons. Because their spectral properties are not available or incomplete, they will be ignored during the subsequent structure generation. The ignored building blocks are displayed in red by default.

II-10 Manual Editing of the Building Blocks

In this step we designate the carbon atoms that are part of those aromatic building blocks and exclude them from the subsequent structure generation. From the chemical shifts, the building blocks #8, #9, #11 - #18 are readily recognized as aromatic ones. Choose Analysis/User-Defined Building Blocks. In the displayed palette, select Modify, and turn on Ignored Atom.

Then click on building blocks #8, #9, #11 - #18, except #15, which is already treated as ignored atom. In this way all of these aromatic atoms are defined as ignored atoms (displayed in red by default), and they will not be included in the subsequent structure generation.

II-11 Interpretation of Bond Constraints

In this step NMR-SAMS interprets the 2D NMR spectral data (except HMQC) as bond constraints, and sets up an atom-atom connection matrix (ACMX) for structure generation. The control parameters for spectral interpretation can be adjusted by the user by selecting the Parameters with the pull-right option NMR Interpretation in the Edit menu. The default values are used here as shown in the following dialog box

The parameters that control the setup of ACMX can be adjusted by selecting the Parameters with the pull-right option Setup ACMX in the Edit menu. The default values are used here.

Select the Bond Constraints in the Analysis menu. NMR-SAMS uses the peak lists in the NMR data file (.nmr file) and saves the results in the master data file (.mdf file). In the following sections the results of NMR data interpretation are described for each type of peak list. All of these actions happen after activating the Bond Constraints option in the Analysis menu.

COSY Peaks:

The bond constraints regarding the correlated ¹H spins are listed in the MDF after a keyword "COSY" according to the interpretation controls specified by the user in the dialog box regarding NMR Interpretation. (This can be accessed by selecting the Parameters file with pull-right option Spectrum Interpretation in the Edit menu. For this example, the default parameters are used). The three potential long-range coupled COSY peaks are interpreted as 3-5 intervening bonds between the correlated protons. The remaining short-range coupled ones are interpreted as 2-3 bonds.

NMR-SAMS also warns the user to avoid some common pitfalls that may lead to incorrect structure generation. For example, if two ¹H peaks are very close and no COSY peak is observed between them, the user is alerted to check if any COSY is neglected between them. If the user is not sure about this, the program allows the user to add a "pseudo bond constraint" on this proton pair just in case that the COSY peaks is neglected. (This threshold for checking near-diagonal COSY peaks can be changed by the user by selecting Parameters with the pull-right option NMR Interpretation in the Edit menu. The default value is 0.02 ppm).

HMBC Peaks :

Each peak is interpreted as a bond constraint of 2 or 3 intervening bonds on the relevant ¹³C and ¹H spins. Here 75 bond constraints are listed in the MDF after a keyword "HMBC".

Transformation of Various BCs into C-C BCs :

The various bond constraints, e.g. the H-H BCs from COSY and the C-H ones from HMBC, are transformed into C-C bond constraints based on the HMQC-derived C-H connectivity. NMR-SAMS also cross checks the bond constraints for mutual consistency, and alerts the user whenever a contradiction is encountered. The bond constraints are written into the .mdf file after the keyword “C13~~C13”.

The format of the bond constraint (BC) is described below. A BC is represented in one line with the following information:

(Atom_y ... - Atom_x ... : minBond ~ maxBond; BondType; minNSBC ~ maxNSBC)Source

where

Atom_y ... is the correlated atom(s) on the Y dimension (¹³C domain for a heteronuclear spectrum). It could be more than one in the case of ambiguity.

Atom_x ... is the correlated atom(s) on the X dimension (¹H domain for a heteronuclear spectrum). It could be more than one in the case of ambiguity.

minBond and maxBond are the minimum and maximum bond separations between the relevant atoms.

BondType is the type of the intervening bond between the atoms: 0, 1, 2, or 3 for unknown, single, double, and triple, respectively.

minNSBC and maxNSBC are the minimum and maximum numbers of relevant atom pair(s) that must be satisfied for this BC.

Source encodes the cross peaks (or other source) from which the BC was derived. A cross peak is represented by its spectral type and its ID number. The following codes are used to represent the spectral types:

“C” for COSY, “Q” for HMQC (or HETCOR), “B” for HMBC (or COLOC), “N” for NOESY, “I” for INADEQUATE.

The following codes are used to represent the other kind of sources:

“S” for a pseudo BC added by the program, “U” for a user-defined BC, and “G” for a previously generated bond (when using a previously generated substructure as the starting point for next structure generation).

For this example (paclitaxel), some information messages are given (as indicated below) when different number of intervening bonds are assigned to the same relevant atoms, or if the relevant atoms of a bond constraint is found to be the sub-set of another bond constraint. Such information can be viewed in the .log file by choosing Log File option from the Edit menu.

When this task is completed a unified set of 62 C-C bond constraints are obtained and listed in the MDF after the keyword “C13~~C13”. One my notice that after such transformation and cross-checking, some of the BCs have more than 10 individual cross peaks. For example, the following C-C bond constraint:

(17 - 23: 1 ~ 1; 0; 1 ~ 1)C10Q8Q13C12Q8Q14B50Q13B52Q14B74Q8

means that there is a bond between C17 and C23, and this is derived from the 12 cross peaks denoted following the right parenthesis, where “C10” denotes COSY peaks #10, “Q8” HMQC peaks #8 and so on. (The numberings of the peaks correspond to those listed in Tables II-V).

During the bond constraint transformation, COSY correlations between geminal protons are automatically discarded. To avoid overlooking a correct structure, if a ¹H peak is found to correlate with multiple ¹³C peaks, this ¹H peak is taken as a degenerate peak and a pseudo bond constraint is added between the corresponding carbons. It is due to the fact that correlation between degenerate proton are generally not observed. In the case of paclitaxel this doesn’t apply because the degenerate peaks due to the aromatic rings are discarded.

Setting up Atom-Atom Connection Matrix (ACMX)

This step performs the task of defining the bonding possibilities between the building blocks based on the available bond constraints. It is automatically done after interpreting the bond constraints. The unambiguous bond constraints (one bond between exactly two atoms) are treated as fixed bonds (by default), and the rest are used as constraints during the subsequent structure generation.

For this example, the program automatically sets up the ACMX and shows the following summary of constraints. Click OK to accept it.

The building blocks, as well as some fixed bonds are displayed as seen below. An atom with unsatisfied valence is marked by a “*” and displayed in different color (blue by default). The ignored carbon atoms are displayed in red by default. The fixed bonds with unknown bond types are displayed as dashed lines. A fixed bond with an unknown bond type can become either a single, double, or triple bond after the subsequent structure generation. The user can choose Display Options in the Display menu to change the displayed features. For example, one can choose to display the associated ¹³C and ¹H chemical shifts of each building block, as well as a connection table listing all the building blocks and the available bond constraints. The ACMX itself is not displayed but can be inspected by choosing Master Data File in the Edit menu.

Note that the program automatically adds a carbonyl and 5 carboxylic groups based on the ¹³C chemical shifts. See the next section about how to edit them.

II-12 User-Defined Bond Constraints

Before submitting the ACMX to NMR-SAMS for structure generation, it is important for the user to verify the building blocks, fixed bonds, and the available bond constraints. For complex molecules, if the user has a priori knowledge about the structure, then additional constraints can be defined in the form of “user-defined bond constraints”. User-defined bond constraints concerning heteroatoms (O, N etc.) derived from other spectral data are especially important for improving the efficiency of the subsequent structure generation (shorter computational time and fewer candidate structures). Normally connectivity information concerning these atoms are not available from 2D NMR spectra, but the user can use additional information such as functional groups from IR, UV or MS data, to define the non-NMR constraints.

NMR-SAMS allows the user to define bond constraints based on a set of building block, or a substructure if structure generation has already been done. For this example, as the structure generation has not been done yet, only the building blocks show above are available.

Next select User-Defined Bond Constraints in the Analysis menu. A User-Defined Bond Constraints editor palette as shown below is displayed. With this editor the user can add or delete bonds between the building blocks. NMR-SAMS checks each added bond against the available fixed bonds and bond constraints and rejects inappropriate ones. After editing, click OK button in the user defined bond constraints editor to set up the ACMX again based with the additional user-defined bond constraints included. This process can be repeated to add more bond constraints or to delete undesired ones.

For paclitaxel, a carbonyl and 5 carboxylic groups were already added by the program automatically. Based on the ¹³C chemical shifts, the five carbons (C-2 to C-6) with chemical shifts ranging from 172.921 to167.237 PPM are either ester or amide groups. In the HMBC spectrum a peak is observed between C-5 or C-6 (because they are partly overlapped) and the NH at (¹³C: 167.2, ¹H: 7.0). This means that either C-5 or C-6 can be an amide group. So C-6 is arbitrarily set as an amide group. To change this, first select Delete, and click C-6 and O-59 to delete the bond between them. Next select Add and Single Bond, then click C-6 and N-48 to add a bond between them. Then click OK to accept the functional groups.

NMR-SAMS automatically creates the ACMX with the updated user-defined bond constraints included. Then a summary of the constraints is displayed again (as shown below). Click OK to accept it.

Then the updated functional groups and fixed bonds are displayed as shown below:

Note in the case of partial structure elucidation, where typically insufficient spectral data is available, the structure generation generally takes more time to converge. Even after convergence it is not always guaranteed to be a correct structure. Therefore any additional user-defined constraints will improve the convergence and the quality of the structure.

II-13 2D Structure Generation

One of the major bottlenecks in computer-assisted structure elucidation is the efficiency of structure generation (which is a factor of the computation time, the quality of structure generated, and the number of candidate structures generated). The structure generator searches all the plausible 2D structures that are consistent with the data. If the spectral data is precise and has fewer ambiguities, then NMR-SAMS usually generates the correct unique structure within reasonable computation time. However, as structure elucidation is combinatorial problem, it becomes more and more time consuming as the size of the molecule and the number of free bonds increases (for example molecules with sizes > 30 heavy atoms). Several heuristic rules are applied to speed up the process of structure elucidation. Some of these controls can be adjusted by the user by selecting the option Parameters with pull-right option 2D Structure Generation in the Edit menu. A dialog box will be displayed (as shown below) and the details of the parameters in the dialog box can be seen by selecting the on-line help.

In this example (paclitaxel), except for the maximum number of violated bond constraints which is set to one (instead of zero) the rest of the parameters are set to default values. If this is not set the NMR-SAMS will not generate any structure. The reason is that the HMBC peak at (¹H: 2.453, ¹³C: 87.71) corresponds to a four-bond coupling between OH-50 and C-9, while all other HMBC peaks correspond to either two- or three- bond couplings. So select Parameters with the pull-right 2D Structure Generation in the Edit menu, and make sure that 1 is the value for Maximum Limit of Bond Constraint Violation. Next click OK to apply the modified parameter.

After choosing Generate 2D Structures in the Analysis menu, the user is prompted to define a range of "dummy bonds" to be fixed in the generated structures (as shown below in the dialog box). For partial structure elucidation (PSE), the structure generator will try to generate the largest substructure which is consistent with the available data. In some instances a dummy bond is used to satisfy a free bond by assuming that it is connected to one of the ignored atoms. In the case of paclitaxel the three phenyl groups are not included in the structure elucidation process, so we type "3 3" to add exactly three dummy bonds in each generated structure.

During the structure generation process, a Structure Generation in Progress dialog box is displayed (as shown below). This dialog box shows the initial state (in terms of the Molecular Formula, Number of free bonds and number of bond constraints used )and the current state (in terms of the number of completed structures) of the problem. The dialog box is updated at a frequency based on the parameter Interval for Updating Structure generation Dialog Box. The default value is 0.1 minute. (Users can change this parameter by selecting Parameters with pull-right option 2D Structure Generation in the Edit menu).

At any time the user can abort the structure generation process by clicking the Stop button. It takes few seconds to update after the Stop button has been selected. The complete structure generation of paclitaxel takes about 15 minutes (CPU time) and 12 candidate structures are obtained. 12 structures are completed in the very first 2 minutes, and thereafter the program continues to search for other substructures which are consistent with the data.

After the completion of structure generation the first structure is displayed along with a structure browser (as shown on the following page) which enables the user to browse through the remaining structures one at a time. All of the candidate structures have three dummy bonds (denoted as "~") which supposedly lead to the ignored atoms (displayed in red by default). These candidates are called “complete structures” because they do not have any remaining free valences, though they are chemically incomplete. In all 38 largest substructures generated are retained and the user is prompted (as shown in the dialog box below) to save them along with the complete ones. Click OK to save them in the structure file (.str file).

By dragging the slider in the Structure Browser, the candidate structures/substructures can be displayed one after another. The candidate structures are displayed with the carbon atoms labeled by their assigned ¹³C and ¹H chemical shifts. By selecting Display Options in the Display menu, the user can also choose to display a connection/assignment table of the structures. The status of different NMR-derived bond constraints, whether satisfied or violated, is summarized in the table.

Alternately, while entering the user-defined bond constraints, if the user had only connected C-5 and C-6 to an oxygen by double bond (because it was not apparent which one was amide or ester group), then the structure generation would take about 2 hours (CPU) to complete and 24 candidate structures will be generated. Among them only 12 are unique ones. Each duplicate pair of candidate structures show the exchange of location of C-5 and C-6. (See data set paclitaxel-test2.*).

It can be seen that most of the multiple candidates arise from the lack of HMBC connectivity between C-2 and H-3 (on C-16), as well as between C-4 and C-10 (both are quaternary carbons). By adjusting experimental parameters or running different experiments one should be able to include additional information.

II-14 Editing Generated Structures

After the structures have been generated, one can further edit the structures to complete the portions that were ignored during the structure elucidation process. These ignored atoms are displayed in red as individual atoms besides the candidate structure. To perform this task, first choose the correct structure to be displayed (#1 in this case) by using the structure browser. Next, select Generated Structures option under the Edit Menu to activate the molecular editor shown in the figure below. When you edit a generated structure, NMR-SAMS will not check the edited structures against the bond constraints, shifts, etc. In the following section, the user is advised to create these phenyl groups and attach them to the carbons marked with "~". The phenyl groups are shown in the figure below.

Click Bond on the Molecular Editor to switch to adding bond mode. Add bonds between the ignored atoms by clicking left mouse button on two atoms and a bond (default: single bond) will be added between them. To stop adding bonds, click the right button on the mouse. To add double bond, select Double on the Molecular Editor panel. In this manner build three phenyl groups with alternating single and double bonds between the ignored atoms. Then connect each phenyl group to the different dummy bonds (the dummy bond will not disappear). Next select the Display Options with pull-right option Refine in the Display menu to refine the display of the edited structure. This represents the complete structure of paclitaxel as shown above.

Next click OK to exit the Molecular Editor palette. Select Export in the File menu with pull-right option Structures (MDL) and the structure will be exported into a file called paclitaxel-test001.mdl in MDL format.

II-15 Exporting NMR Data, 2D Structures, and Resonance Assignments

NMR-SAMS provides a tool to export 2D structures to third party molecular drawing programs such as Chemdraw, ChemSketch etc. The structure is exported with coordinates in MDL format. To export structures, select Export option with the pull-right option Structure (MDL) in the File menu, and the current displayed structure will be exported into a file paclitaxel-test00x.mdl, where x is the sequential number of the structure.

NMR-SAMS also provides export functions for NMR peak lists (in the form of chemical shift correlations) and resonance assignment results. To create chemical shift correlation table, Select Export with pull-right option Chemical Shift Correlations in the File menu. The correlation of chemical shifts are written into a file (paclitaxel-test.spc) in a format that is familiar to the chemists.

To create assignment table, select the option Export with pull-right option Assignment in the file menu. The resonance assignments of the candidate structure or substructure which is on display will be written to a text file paclitaxel-test00x.rst, where x is the sequential number of the (sub)structure. This file contains the ¹³C and ¹H assignments of all the assigned atoms in the molecule. If NOESY peaks are available, then the assignment of the NOESY peaks along with distance constraints and the actual bond separation between the relevant protons are included. This information enables resolving of ambiguous NOE peaks, and identification of through-space NOE connectivities.

II-16 Structure Elucidation With Unknown Molecular Formula

If you do not know the molecular formula, you can still do the structure elucidation using NMR-SAMS. An example of this is shown in the working data set taxol-nomf. The procedure is mostly the same as in the case where molecular formula is known. The main differences are described here:

1. While creating the data set, input molecular formula as ‘unknown’ (see II-3).

2. During the peaks table conversions (see II-4-II-8), the program will not check the results against the molecular formula. So some of the warning messages will not appear.

3. During the generation of building blocks, it will generate building blocks according to the observed ¹³C peaks. So for this example, fewer building blocks will be generated compared with those obtained in section II-9:

4. During Manual Editing of Building Blocks (see II-10), you have to add more ignored building blocks that correspond to the phenyl groups. First modify building blocks #8, #9, #11 - #18 as ignored atoms. As shown below, select Add in the palette, type “O” as Element, turn off Ignored Atom, make sure Proton Count is 0 and Valence is 2. Next click in the main graphics window 11 times to add 11 oxygen atoms. Next click in the main graphics window 11 times to add 11 oxygens.

Next, as shown below, keep Add selected in the palette, type “C” as Element (Ignored Atom is automatically selected), select Unknown as Proton Count, and make sure Valence is 4. Next click in the main graphics window 8 times to add 8 carbon atoms with uncertain number of attached protons. These carbons correspond to those for which ¹³C peaks are not observed.

After clicking OK in the palette, NMR-SAMS rearranges the building blocks and automatically sets up the ACMX again. The resulting building blocks should be the same as the final results described in II-10.

For the rest of the steps (following Section II-10), the operations and results should be exactly the same as those described for the example with a known molecular formula.

Computer-Assisted Structure Elucidation of Paclitaxel (Taxol) Using SpecMan and NMR-SAMS

Table of Contents

Computer-Assisted Structure Elucidation of Paclitaxel (Taxol) Using SpecMan and NMR-SAMS