ScienceSoft, LLC
www.ScienceSoft.net
TOC PREV NEXT INDEX


CHAPTER 5: Tutorial III: Combining Analysis Results

CHAPTER 3: "Tutorial I: Using NMRanalyst" and CHAPTER 4: "Tutorial II: Setting Analysis Parameters" use 2D INADEQUATE for the molecular carbon skeleton determination. Indirect detection NMR can improve the detection sensitivity over an order of magnitude. The graphical illustration shows a carbon-proton bond is observable through an HSQC spectrum. Combining this information with 2-bond and 3-bond HMBC correlations provides carbon-carbon bond information. The AssembleIt workwindow combines the analysis results from different spectrum types. It then derives the most likely molecular skeleton(s) from this often incomplete and ambiguous information.

Dr. Péter Sándor (Varian Deutschland GmbH) graciously provided 500 MHz NMR datasets of gibberellic acid (MF: C19H22O6).1 Using a 7 mg/ml sample in DMSO-d6, a 1D carbon (acquisition time: 3 h 37 min), 1D proton (14 s), multiplicity edited gHSQC (31 min), and absolute value gHMBC (2 h 31 min) were acquired. Similar to the CHAPTER 4: "Tutorial II: Setting Analysis Parameters", create a copy of the supplied gibberellic_acid dataset containing the experimental data (FIDs). Delete its HMBC and HSQC subdirectories to start with default analysis parameters. In NMRanalyst, start [Edit] [Preferences...], set Mode: to [Full NMRanalyst], deselect the [Show All Input Fields] switch, and click [OK].

5.1 Analysis of 1D Spectra

NMR-based structure elucidations start with the analysis of 1D survey spectra. The proton spectrum is fast to acquire and is regularly used. Carbon and nitrogen 1D spectra are less sensitive to acquire. NMRanalyst supports deriving their shifts indirectly as described later in this chapter.

From the NMRanalyst application window, click the [Directory Editor] button. Set NMRDATA to the copied gibberellic_acid directory, set NMRSPEC to HMBC, and click [OK]. Switch to the 1D Analysis workwindow. Load from directory carbon.fid the spectrum specific parameter file procpar. Run the workwindow.

From the Graphic workwindow display the created and shown carbon.plot file. All resonances are detected as shown by their black color. The strong resonances around 40 ppm belong to the DMSO-d6 solvent. In the 1D Analysis workwindow, load from directory HMBC the file 1d_analysis.log. This procedure loads the reference corrected carbon resonances in the input screen and selects the [Remove] switches for the seven DMSO-d6 resonances. Rerun the workwindow.

From the Spectrum Type menu, select [HSQC]. Switch to the 1D Analysis workwindow. Load the spectrum specific parameters from directory proton.fid file procpar. Run the analysis of the 1D proton data and display the shown proton.plot file.

5.2 Analysis of Multiplicity Edited HSQC Spectrum

For the gibberellic acid structure elucidation, switch to the FFT workwindow. Set the acquisition specific parameters by loading from directory ghsqc.fid the file procpar. The 1D carbon spectrum was referenced by -93.722 Hz, so update this F1: Start of Spectrum value to 954.79 Hz. Run the FFT workwindow. Change to the 2D Analysis workwindow. Run the 2D Analysis and the Report workwindow is auto-run.

The next step is to load and lock the phase functions to distinguish CH, CH2, and CH3 groups from this multiplicity edited HSQC spectrum. In the Graphic workwindow load the HSQC/report.log containing the determined linear phase functions, so the following displayed 2D HSQC is phased. Also load these phase functions in the 2D Analysis workwindow to improve the analysis results. Run the 2D Analysis and it auto-runs the Report workwindow.

In the Graphic workwindow, set File With Plot Data to hsqc.spec and display this spectrum by starting this workwindow. Typically 2D INADEQUATE (described in CHAPTER 4: "Tutorial II: Setting Analysis Parameters") and 1,1-ADEQUATE spectra (described in CHAPTER 6: "Tutorial IV: Advanced Structure Elucidation") have too low a signal-to-noise ratio to create meaningful spectral projections. But for the other spectrum types, 1D projections of a 2D spectrum are shown by default. Were all the correlations detected? The undetected resonance around 40 ppm in F1 and 2.5 ppm in F2 arises from partly undeuterated DMSO-d6 solvent and should not be detected. The F1 43 ppm and F2 5.1 ppm resonance is too far off-diagonal and should not have been detected. The Report workwindow output screen suggests how to adjust the integral threshold to avoid such problems:

 HSQC Resonances of Wrong Sign
 -----------------------------
 #   721  C 18  H 42  F1=  16.579  F2= 1.827  I,max=-353.832  I=  11.675
 => Recommended Threshold: Integral = 11.676

 Positive HSQC Resonances too Far From Strongest One for Same Carbon
 -------------------------------------------------------------------
 #   259  C  9  H 21  F1=  68.513  F2= 3.535  Fi,max= 3.858  I=  44.059
 #   784  C 19  H 56  F1=  14.569  F2= 0.986  Fi,max= 1.061  I=  56.504
 => Recommended Threshold: Integral = 56.505

 HSQC Resonance Much Weaker Than Strongest One for Same Carbon
 -------------------------------------------------------------
 #   596  C 16  H 15  F1=  42.711  F2= 5.115  I,max= 371.826  I=  11.223
 #   718  C 18  H 39  F1=  16.532  F2= 1.860  I,max= 353.832  I=  13.844
 => Recommended Threshold: Integral = 56.505

Set the 2D Analysis Thresholds: Integral field to 56.505 as suggested. Rerun the 2D Analysis which auto-runs the Report workwindow. Display the shown structure.plot file of HSQC derived molecular structural results. The multiplicity editing of the HSQC inverts CH2 relative to CH3 and CH resonances. Identified CH carbons are shown in green, CH2 ones in blue, and CH3 ones and proton resonance frequencies in orange.


5.3 Analysis of Gradient HMBC Spectrum

Switch to the HMBC spectrum type and display the FFT workwindow. Load from directory ghmbc_12Hz.fid the procpar file. The 1D carbon spectrum was referenced by -93.722 Hz, so update this F1: Start of Spectrum to 879.678 Hz. Run the FFT workwindow. Switch to the 2D Analysis workwindow. Run the 2D Analysis which auto-runs the Report workwindow.

For this "absolute-value" HMBC, only two of four 2D phase components were acquired. Proton-proton couplings during the HMBC pulse sequence cause F2 phases to scatter over the 2 phase range and no linear phase function can be derived. For HMBC, don't set phase functions so NMRanalyst determines individual F1 and F2 phases for every correlation.

Display the shown integrals.plot file of the correlation integral distribution. To avoid the report of longer-range correlations, set the 2D Analysis workwindow Thresholds: Integral to 0.26. Rerun the 2D Analysis which auto-runs the Report workwindow.

From the Graphic workwindow, display the shown HMBC spectrum (set File With Plot Data to hmbc.spec). As for the HSQC spectrum before, check if the strong resonances were detected as indicated by rectangular bounding boxes. The pairs of signals close to the spectral diagonal, 100 to 200 Hz apart, are incompletely suppressed direct C-H bond correlations. Weak correlations are likely 4-bond or longer-range correlations which are not needed for the structure elucidation.

5.4 AssembleIt: HSQC & HMBC Derived Gibberellic Acid Structure

Hetero nuclear 2D NMR spectra can be acquired of samples an order of magnitude smaller than for the direct carbon skeleton determination through 2D INADEQUATE. AssembleIt is our approach to solving the resulting puzzle of deriving the molecular carbon skeleton. The detected hetero nuclear information can be incomplete or ambiguous. NMRanalyst allows up to a specified number of experimental correlations to be 4-bond or even to be incorrect assignments and the best possible structures can still be derived.

Display the AssembleIt workwindow. It first combines results from analyzed spectrum types. Specified non-existing input files in the COMBINE NMR ANALYSIS RESULTS section cause a warning and are ignored. From the row of switches, select Consider: [Chemical Shift Rules] to identify likely double bonds and the location of carbonyl groups. Also select [Ambiguities] to consider correlations with assignment ambiguities. (Resulting carbon-carbon correlations are stored in the file specified in the Output: Atom-Atom Correlations field, which is set to CC_corrs.plot by default. The input field is displayed when the [Edit] [Preferences...] [Show All Input Fields] switch is selected).

AssembleIt derives the most likely molecular skeleton(s) from the above derived correlations. It generates up to the number of requested most likely structures. Most important for building structures are 2-bond HMBC correlations. Some are not detected. For the gibberellic acid structure generation, select the AssembleIt: ELUCIDATE MOLECULAR STRUCTURE FROM NMR DATA section. Set its Evaluate: Unobserved Bonds to 7 to have AssembleIt derive up to this number of unobserved bonds from other detected correlations. Run the AssembleIt workwindow. The exhaustive structure generation takes under 1 minute (on a 3.06 GHz PC) and one structure results.

From the Graphic workwindow display the shown CC_corrs.plot file first. The colored labels in this circle diagram represent the 19 gibberellic acid carbon shifts. A red atom color identifies a quaternary, green a CH, blue a CH2, and orange a CH3 group. The selected Consider: [Chemical Shift Rules] switch further reduces the number of free valences of some carbons. This is indicated by an appended atom suffix of s, d, t, or q. Singlet (unprotonated), doublet (one bonded proton), triplet (two bonded protons), or quadruplet (three bonded protons) represents the number of corresponding resonances which would be observed in a proton-coupled carbon spectrum. The dotted lines in this circle diagram represent the derived unambiguous correlations and the dashed lines the ambiguous ones.

Display the resulting structure, file CC_corrs.plot.1. In this case, only the correct gibberellic acid carbon skeleton is consistent with all the experimental correlations. In this display, bonds which cannot contain an unobserved heteroatom are shown as solid lines. The other bonds are shown in a dotted line style. Bonds which are not observed, but were derived from longer-range correlations, are identified by a question mark label. The other bond labels show the proton frequency over which the bond was derived. The direction over which a bond is observed is indicated by arrows.

5.5 Structure Identification Without 1D Carbon Spectrum

Acquiring a 1D carbon spectrum for the gibberellic acid structure identification or elucidation sounds harmless. But its 3 h 37 min acquisition time exceeds the remaining 1D proton, multiplicity edited HSQC, and HMBC acquisition times combined. This section describes generating a carbon resonance list from the HSQC and HMBC spectra. From the current gibberellic_acid directory, remove the carbon FID and related information. Start the UNIX Shell and issue the command:

 % rm -r carbon.*

With selected HMBC spectrum type, display the 1D Analysis workwindow. Set Input File Format: to [Generate Generic List] to generate an initial carbon resonance list. Copy the FFT workwindow F1 settings into the corresponding 1D Analysis input fields: Observe Frequency of 125.704 MHz, Spectral Width of 22522.523 Hz, Start of Spectrum of 879.678 Hz,2 and Number of Points of 1024. Click [Start]. The resulting 254 carbon resonance frequencies are 88.0646 Hz apart covering the whole HMBC F1 carbon range. They are saved in the specified file carbon.out.

Now analyze the HSQC spectrum with carbon.out: Switch to the HSQC spectrum type. Its 1D Analysis and FFT workwindows are unaffected. Switch to the 2D Analysis workwindow. The F1 1D Analysis Output File uses the carbon.out generated list. To cover its resonance spacing, set Map F1 Frequencies ± to 45 Hz (half the resonance spacing). Increase Thresholds: Integral to 70, as the extensive mapping creates slightly larger values than before. Select the [F1 Phase] and [F2 Phase] switches in the Mapping and Detection sections. Remove both phase functions (F1 Phase = and F2 Phase = fields) as they were determined with the help of the 1D carbon spectrum. In the Report workwindow, specify the Redetermined F1 Resonance List as carbon_hsqc.out. Run the 2D Analysis workwindow and it auto-runs the Report workwindow. Twelve carbon resonances are stored in carbon_hsqc.out.

Switch to the HMBC spectrum type and its 2D Analysis workwindow. The F1 1D Analysis Output File is set to the generic carbon.out. Set Map F1 Frequencies ± to 45 Hz to cover the resonance spacing in this generic list. Set Thresholds: Integral to 1.1 and Agreement to 30. In the Report workwindow set Redetermined F1 Resonance List to carbon_hmbc.out. Run the 2D Analysis workwindow and it auto-runs the Report workwindow. This exhaustive HMBC analysis takes about 15 minutes (on a 3.06 GHz PC), but is much faster than acquiring a carbon spectrum. Fourteen identified carbon resonances are stored in carbon_hmbc.out. Now the combined carbon line list can be generated. From the UNIX Shell window execute:

 % merge2list carbon_hsqc.out carbon_hmbc.out 0.12 > ncarbon.out

This merges the both line lists, keeping HSQC determined carbon resonances and adding the HMBC ones over 0.12 ppm apart from other resonances. The resulting 19 resonances are saved in the file ncarbon.out.

Select the HSQC spectrum type and switch to the AssembleIt workwindow. Select the [FindIt: IDENTIFY DATABASE STRUCTURES BEST MATCHING NMR DATA] switch. In the [COMBINE NMR ANALYSIS RESULTS] section, set Carbon to the generated ncarbon.out list. Start this workwindow. The best matching structure identified has the PubChem CID of 6466 and is gibberellic acid.

5.6 Structure Elucidation Without 1D Carbon Spectrum

With the previously generated ncarbon.out, the gibberellic acid structure elucidation can be completed without using the acquisition time intensive 1D carbon spectrum. Repeat the HSQC and HMBC spectrum analysis using ncarbon.out: Select the HSQC spectrum type and its 2D Analysis workwindow. Set F1 1D Analysis Output File to ncarbon.out. Delete the Map F1 Frequencies ± entry. In the Report workwindow, delete the Redetermined F1 Resonance List entry. Run the 2D Analysis workwindow again and it runs the Report workwindow. Load the determined phase functions in the Graphic and 2D Analysis workwindows. Run the 2D Analysis workwindow and the Report workwindow is auto-run.

Switch to the HMBC spectrum type and its 2D Analysis workwindow. Set F1 1D Analysis Output File to ncarbon.out, Thresholds: Integral to 0.26, and delete the Map F1 Frequencies ± and Thresholds: Agreement entries. In the Report workwindow, remove the Redetermined F1 Resonance List entry. Run the 2D Analysis and it runs the Report workwindow. In the AssembleIt workwindow, set the input field Carbon to ncarbon.out. Set in the AssembleIt section the Evaluate: Unobserved Bonds to 7, 4-Bond HMBC Correlations to 1, and start this workwindow. One potential gibberellic acid structure is generated in 5 minutes (on a 3.06 GHz PC) and it is the correct one.

Extracting carbon list information from the 2D spectra requires more user interaction than analyzing a 1D carbon spectrum. But it is faster than acquiring a 1D carbon spectrum. It potentially allows using smaller sample quantities or shorter total acquisition times. As described in SECTION 2.2: "Contents of the NMRanalyst Distribution", the NMRanalyst distribution contains the ca. 200 micro-gram dihydrotestosterone FIDs. For under one milligram of compound, the acquisition of a 1D carbon spectrum is practically impossible. But its structure elucidation succeeds with the techniques of this chapter. See http://www.sciencesoft.net/dihydrotestosterone/index.html for details.

A limitation is the required separation among carbons to be distinguished. Deriving quaternary carbons is done from HMBC. Due to the possible F1 skew of HMBC resonances, carbons need to be separated 0.12 ppm (this is also the default value) from other carbons for reliable distinction. Protonated carbons can be determined from the phase sensitive HSQC. Here the HSQC F1 resolution is important. Carbons down to a fifth of this resolution are distinguishable by the software, which means down to thirteen Hz for the gibberellic acid HSQC. Incompletely resolved multiplets in F2, or closely spaced carbons with bonded protons at the same F2 frequency further limit this distinction.

For analyzing your own datasets, work at least through CHAPTER 6: "Tutorial IV: Advanced Structure Elucidation". As described in this section and CHAPTER 6: "Tutorial IV: Advanced Structure Elucidation", the 4-Bonds HMBC Correlation and Long-Range HMBC Correlation fields might need to be specified. Also, the AssembleIt workwindow Weak threshold should be specified to accelerate the structure generation. The FindIt structure identification reliably finds similar structures. Full AssembleIt structure elucidations are more likely to encounter unexpected challenges making them less reliably than FindIt applications.

1The plant growth regulator gibberellic acid (GA) is a potent plant hormone. Its natural occurrence controls the plant development. In gardening, GA is used for seed germination, growth promoter, early flowering, and increasing the yield of greenhouse tomatoes.

2The carbon dimension was referenced based on the 1D spectrum. So strictly, the original referencing should be used. But either approach works.



ScienceSoft, LLC
www.ScienceSoft.net
TOC PREV NEXT INDEX