NUS Reconstruction of 4D spectra in Topspin

Table of contents

• General workflow
  • Troubleshooting
    • Post-reconstruction phasing
    • The reconstructed spectrum is left in the time domain
• Examples
  • Bruker 4D HCNH NOESY hsqcnoesyhsqccngp4d
    • Processing steps

  • [AI	ffinity custom 4D HMQC-NOESY-HSQC experiment noehcnhwg4d_nove](#aiffinity-custom-4d-hmqc-noesy-hsqc-experiment-noehcnhwg4d_nove)

    • Processing steps
  • Frank Lohr’s implementation of 4D HMQC-NOESY-HSQC experiment sfhmqc_noe_sfhmqc_4Dhcnh.fl
    • Processing steps
• Credits

Tested environment:

• Topspin 4.4.0 with commercial license for NUS processing
• OS: AlmaLinux 9 (as an Oracle virtual machine)
• 20 CPU cores, 120 GB of RAM.

[!WARNING] Not everything is thoroughly tested yet

General workflow

1. Copy the raw 4D spectrum in a new directory by executing wrpa command. This will make working with the 4D neat and safe.

   [!NOTE] NUS FIDs before reconstruction are not that heavy: the ser files are usually <1 Gb.

2. Switch to the newly copied procedure; Open up the processing parameters (edp or the tab PROCPAR).
3. Adjust the size of the dimensions in the final spectrum with the parameter SI. Always set it to a power of 2. For example, if TD values under ACQUPARS is 160, then se SI to 256 or 512.

   [!IMPORTANT] Always set the SI to the next or higher power of 2, never lower than the respective TD value! Otherwise TopSpin behaves weirdly, e.g. will leave indirect dimensions in time domain.

4. If there were recorded test 2D planes (or 3D cubes) with this exact 4D pulse program, check their processing parameters:
   • Phasing in the direct and indirect dimensions
   • Window functions (rarely changes from default)
   • Baseline correction parameters (rarely changes from default)
   • Linear Prediction (LP), if you use it.

     [!TIP] NUS and LP should not be combined. NUS simulates fitfully the whole FID, while LP simulates the FID decay that was truncated out. As such, NUS reconstruction substitutes LP and LP must not be applied to any spectrum (4D, 3D) recorded with NUS. If recorderd without NUS, Using LP can enhance resolution, especially if the time domain (TD) values are small or if your FIDs are truncated. In TopSpin, you can use LP for improving resolution in particular dimensions during the Fourier Transform process by specifying the ME_mod and NCOEF parameters for those dimensions. Note that it may cause additional wiggles.

1. Set the FnMODE according to the acquisition mode (usually the N dimension is echo-antiecho and the rest States-TPPI)
2. Note down the signal region in the direct dimension. It can be extracted either from the test planes or the 2D experiments (HSQC, TOCSY, etc).
   • Go to the 2D experiment. Zoom in such that the signal-free regions are trimmed as much as possible. Issue the .ftf2region STSR command, it will prompt to Save display region to Parameters STSR\SI.
   • Issue the STSR command. Note down the values for the direct dimension. Same with STSI

     [!TIP] Those values may be obtained manually: in the 1D or 2D spectrum, note the “col Index” (16 on the screenshot). Save it into STSR. Move the coursor to the right and calculate the width of the dimension in points, save that number into STSI.

[!IMPORTANT] If you copy the FT region from the planes, make sure the spectrum windows (SW) of the 4D and the test planes are the same. If they are not, the signal regions have to be adjusted manually.

1. Go to the NUS section. Set the NUS mode to cs. Set the phasing of the indirect dimensions to the same values as in the Phase section (i.e. PH0 and PH1).
2. For the CS reconstructions in Topspin, it’s a good idea to increase the number of iterations. Since 4D spectra have a high dynamic range, weaker peaks are typically reconstructed in later iterations. The default setting of Mdd_CsNITER 200 (hidden parameter) in Topspin could be increased to Mdd_CsNITER 400 or Mdd_CsNITER 600 for better results. However, this should be tested case by case. While you may be tempted to go up to 1000 iterations, this is likely unnecessary.
3. When everything is ready, issue

`ftnd 0 nusthreads 16`

• ftnd will run the NUS reconstruction, followed by FT all directions, with the Window Multiplication (WM) and baseline correction as specified in the PROCPARS. 0 stands for “all dimensions”, nusthreads 16 allocates 16 CPU cores for the process (Topspin’s limit).

  [!NOTE] Whereas NUS reconstruction is parallelized, FT stage uses only a single thread, therefore takes multiple hours.

  1. Optionally: adjust baseline correction parameters and apply the automatic baseline correction to the whole reconstructed 4D spectrum by issuing absnd.
  2. Evaluate the quality of the reconstructed spectrum by looking at the sum projections. You need to look at both positive and negative projections to identify the antiphase signals as well as the potentially misphased peaks. This is done by the command projplp and projpln. Run each and provide the parameters over the GUI or run the single lines such as projplp 12 all all 21

  • projplp stands for positive projection; run projpln to get the negative one
  • 12 refers to keeping the first two dimensions (for standard Bruker HSQC NOESY, those are C and HC dimensions); that means, N (F3) and direct H (F4) will be summed up.
  • all indicates that all planes within these dimensions should be included.
  • 21 specifies the output PROCNO where the projection data will be stored. Adjust the PROCNO based on where you want to save the output.
    1. Perform automatic baseline correction of the projections with abs1 followed by abs2.
    2. If the projections look good, evaluate the 4D spectrum.
  • Perform a putative peak picking. For that, lower the countours such that the noise disappears, then issue pp and set the sensitivity to the lowest countour level.

  

  • Jump to some position of the 4D which contains signals. For that, press the E button and give the plane numbers or the desired ppm.

  

1. If the phasing or other processing parameters are off, adjust them and repeat the NUS reconstruction. If the projections look good - congratulations!

Troubleshooting

Post-reconstruction phasing

If the spectrum appears to be misphased, it has to be processed all over again. To determine the optimal phasing angles:

The reconstructed spectrum is left in the time domain

• Run ftnd 3, ftnd 2 and ftnd 1 to reconstruct the dimensions one by one.
  • This should not initiate NUS reconstruction, at least on the machines without the NUS license.

Examples

Bruker 4D HCNH NOESY hsqcnoesyhsqccngp4d

The test protein is Ubiquitin measured on the 600 MHz spectrometer.

Download the raw 4D spectrum (Topspin directory of the experiment)

Download the 13C HSQC

Download the 15N HSQC

Axis order:

F4	F3	F2	F1
HN	N	Hc	C

Processing steps

1. Set Main Processing Parameters (edp):

   	F4	F3	F2	F1
   SI	1024	256	512	256
   PHC0	150	0	0	0
   PHC1	0	0	0	0
   PH mod	pk	pk	pk	pk
   BC_mod	qfil or qpol	no	no	no
   STSR	55	0	0	0
   SRSI	350	0	0	0
   FCOR	0.5	0.5	0.5	0.5
   NUS
   Mdd_mod	cs
   MddF180		false	false	false
   MdPHASE		0	0	0

Optionally increase the number of iterations with Mdd_CsNITER 600.

The following parameters are automatically set:

• MDD_CsAlg: IST
• MDD_CsVE: true

1. Process 4D Spectrum: ftnd 0

2. Optional Baseline Correction in F4: absnd 4

3. Correct shift of HC(F2) axis:
   • In Topspin: set SR (F2) to (11.9037/4)*600.05=1785.70 Hz
   • In POKY: set HC shift in st to -(11.9037/4)=-2.976 ppm
4. Pick the most intense peaks with pp interface
5. Make the projections:
   • projplp 34 all all 43
   • projpln 34 all all 430
   • projplp 24 all all 42
   • projpln 24 all all 420
   • projplp 14 all all 41
   • projpln 14 all all 410
   • projplp 23 all all 32
   • projpln 23 all all 320
   • projplp 13 all all 31
   • projpln 13 all all 310
   • projplp 12 all all 21
   • projpln 12 all all 210
6. Evaluate the projections and the spectrum visually by jumping from peak to peak

AI|ffinity custom 4D HMQC-NOESY-HSQC experiment noehcnhwg4d_nove

The test protein is Ubiquitin measured on the 600 MHz spectrometer.

Download the raw 4D spectrum (Topspin directory of the experiment)

Download the 15N HSQC

Download the 13C HSQC

Axis order:

F4	F3	F2	F1
HN	N	C	Hc

Processing steps

1. Correct 1-Point Delay in 13C:
   • Create a new nuslist with a 1-point shift in the 13C axis and replace the original nuslist:

     awk '{print $1,$2+1,$3}' nuslist > nuslist.off
     mv nuslist nuslist.back
     mv nuslist.off nuslist
     

   • Extend the 13C axis in Topspin by 1 complex point by changing static param NusTD(F2) from 160 (TD(F2) in your ACQPARS) to 162 (TD(F2)+2):

     2s NusTD 162
     

     Verify that 162 was saved in the file: grep "NusTD" acqu3s

     This should automatically set static TDeff(F2) also to 162 and allow processing with the adjusted nuslist.

2. Set Main Processing Parameters (edp):

   	F4	F3	F2	F1
   SI	1024	256	256	512
   PHC0	-107	0	0	0
   PHC1	0	0	0	0
   PH mod	pk	pk	pk	pk
   BC_mod	qfil or qpol	no	no	no
   STSR	55	0	0	0
   SRSI	350	0	0	0
   FCOR	0.5	0.5	0.5	0.5
   NUS
   Mdd_mod	cs
   MddF180		false	false	false
   MdPHASE		0	0	0

Optionally increase the number of iterations with Mdd_CsNITER 600.

The following parameters are automatically set:

• MDD_CsAlg: IST
• MDD_CsVE: true

1. Process 4D Spectrum: ftnd 0

2. Optional Baseline Correction in F4: absnd 4

3. Correct Known TopSpin Bug (Badly Written CAR) + Correct Shift by 1/4*SW in 13C:**
   • CAR of HC Axis: Written as 6.666 ppm, but should be 4.7 ppm
     • Total shift of HC(F1) axis:
       • In Topspin: Set SR(F1) to (6.666-4.7)*600.05 = 1179.698 Hz or
       • In POKY: Set HC shift in “st” to -(6.666-4.7) = -1.966 ppm
   • CAR of 13C Axis: Written as 39.1096 ppm, but should be 41 ppm. 13C axis is folded by 1/4*SW, relevant peaks are aliased.
     • Total shift of 13C(F2) axis:
       • In Topspin: Set SR(F2) to (58.0333/4 + (41-39.1096))*150.882693 = 2474.283 Hz or
       • In POKY: Set 13C shift in “st” to -(58.0333/4 + (41-39.1096)) = -16.398 ppm
4. Pick the most intense peaks with pp interface
5. Make the projections:
   • projplp 34 all all 43
   • projpln 34 all all 430
   • projplp 24 all all 42
   • projpln 24 all all 420
   • projplp 14 all all 41
   • projpln 14 all all 410
   • projplp 23 all all 32
   • projpln 23 all all 320
   • projplp 13 all all 31
   • projpln 13 all all 310
   • projplp 12 all all 21
   • projpln 12 all all 210
6. Evaluate the projections and the spectrum visually by jumping from peak to peak

Frank Lohr’s implementation of 4D SOFAST-HMQC-NOESY-HSQC experiment sfhmqc_noe_sfhmqc_4Dhcnh.fl

The test protein is SAK 42D (15.58 kDa) measured on the 950 MHz spectrometer.

Download the raw 4D spectrum (Topspin directory of the experiment)

Download the 15N HSQC

Download the 13C HSQC

Axis order:

F4	F3	F2	F1
HN	N	C	Hc

On the example of the Exp. 72 (Ubiquitin)

Processing steps

1. Set Main Processing Parameters (edp):

   	F4	F3	F2	F1
   SI	1024	128	256	256
   PHC0	-68	90	90	-45
   PHC1	0	-180	-180	0
   PH mod	pk	pk	pk	pk
   BC_mod	no	no	no	no
   STSR	0	0	0	0
   SRSI	400	0	0	0
   FCOR	0.5	1.0	1.0	0.5
   NUS
   Mdd_mod	cs
   MddF180		true	true	false
   MdPHASE		90	90	-45

[!NOTE] This specific phasing in the indirect dimension takes care of the 1-point delay incorporated into the pulse program in ¹³C channel.

Optionally increase the number of iterations with Mdd_CsNITER 600.

The following parameters are automatically set:

• MDD_CsAlg: IST
• MDD_CsVE: true

1. Process 4D Spectrum: ftnd 0

2. Optional Baseline Correction in F4: absnd 4

3. Correct Shift of HC(F1) Axis:
   • In Topspin: Set SR(F1) to (O1P-CNST16)BF1 = (4.7-2.75)950.37 = 1853.22 Hz or
   • In POKY: Set HC shift in “st” to -(O1P-CNST16) = -(4.7-2.75) = -1.95 ppm

4. Pick the most intense peaks with pp interface

5. Make the projections:
   • projplp 34 all all 43
   • projpln 34 all all 430
   • projplp 24 all all 42
   • projpln 24 all all 420
   • projplp 14 all all 41
   • projpln 14 all all 410
   • projplp 23 all all 32
   • projpln 23 all all 320
   • projplp 13 all all 31
   • projpln 13 all all 310
   • projplp 12 all all 21
   • projpln 12 all all 210
6. Evaluate the projections and the spectrum visually by jumping from peak to peak

<!—

Simple alerts

[!NOTE] This is a note.

[!TIP] This is a tip. (Supported since 14 Nov 2023)

[!IMPORTANT] Crutial information comes here.

[!CAUTION] Negative potential consequences of an action. (Supported since 14 Nov 2023)

[!WARNING] Critical content comes here.

---

Authors

• Petr Padrta, 14.6.2024
• Thomas Evangelidis @tevang2
• Ekaterina Burakova @ebur-nmr