リム傑威
Lim Jackwee
Non-Uniform Sampling (NUS) NMR
This is a walkthrough to setting up and processing 2D and 3D NUS pulse programs using method originally developed by the Wagner's group at Harvard (http://gwagner.med.harvard.edu/intranet/hmsIST). The Wagner's homepage also teaches NUS-conversion of pulse program (click). Do note that Topspin 2.0 syntax is needed if you are running NUS pulse programs on Topspin 2.0 instead of 3.0 or latest. The beauty of NUS NMR is the enormous time-s(h)aving to collect the same dataset or even further enhance weak signals in TROSY pulse programs.
If using the istHMS program, please cite the Wagner's Group at J Biomol NMR 2012 52, 315.
Figure from Topspin 3.1.0: http://www.mn.uio.no/kjemi/english/research/about/infrastructure/nmr/manuals/miscellaneous/nus.pdf
Softwares needed
1. nmrPipe/nmrDraw (http://spin.niddk.nih.gov/NMRPipe/install/)
2. FFTW3.3.3 (http://fftw.org/download.html)
3. FFTW3 dynamic library (libfftw3.3.dylib) if not in the FFTW3.3.3 installation package
4. istHMSlinux or istHMSmac (for reconstruction by iterative soft thresholding (ist), from Wagner/ Wand's group or me) (click for redistribution)
5. PoissonGap2.jar (for creating sinusoidally weighted Poisson-gap schedule from Wagner/ Wand's group or me)
6. phf2pipe (for converting 3D or 4D istHMS output into pipe format, from Wagner/ Wand's group or me)
7. parallel (from Wagner/ Wand's group or me)
bruker > fid.com > ft1xyz.com > ft1.com > ist.csh > run.local > phf2pipe > ft23.com > xyz2pipe > pipe2ucsf
2D
In this non-uniform sampling, the Poisson Gap Method is used to create a schedule of sparsely-collected points (Hyberts et al, JACS, 2010).
For a 2D experiment such as NUS-1H-15N HSQC, here we have setup a sampling density of ~10% for number of point, np (-xN) as 96.
In addition, the tolerance level is set to very low e.g. 0.00005 and the sinusoidal weight to 2 for a more random sampling schedule (and with more points collected before evolution decay, unlike 1) and instead of 0 which creates a more uniform sampling schedule. The Randomize output is set to No. However, Yes would allow you to reconstruct even after a few hypercomplex points acquition. Anyway, we will stick to No.
Figure. GUI of PoissonGapUI.jar to create a Poisson-Gap Sampling Schedule and the 9-point sched2d list generated from 10% of 96 points
Once the variables are updated, click execute to generate the sched2d list with 9 sparsely-collected points which will be exported as nus_vclist_t1_2d. In the Topspin Bruker Program, open ajw.nus_hsqc pulse programs and enter -yN as 18 or 2*9 points as in the vclist with 9 points. This will be different from the 3D setup below.
3D
In this non-uniform sampling, the Poisson Gap Method is used to create a schedule of sparsely-collected points (Hyberts et al, JACS, 2010).
For a 3D experiment such as NUS-1H-13C CBCACONH, we have setup a sampling density of ~10% and number of total points, np (-xN) as 180. In addition, the sinusoidal weight to 2 for a more random sampling schedule (and with more points collected before evolution decay) instead of 0 which creates a more uniform sampling schedule. The Randomize output is set to No. However, Yes would allow you to reconstruct even after a few hypercomplex points acquition. Anyway, we will stick to No.
For a typical triple resonance experiment, the 1st indirect dimension is the number of total points for 15N and the 2nd indirect dimension to be the number of total points for 13C. Note that -xN, -yN, -zN is the total points (Real + Imaginary) = 2 x complex point (-xT, -yT, -zT).
For example,
CBCACONH @ 750MHz
From acqus
NUC1/2/3 O1/2/3 SFO1/2/3 Center (O/SFO, ppm) TD (=-xN, -zN, -yN) -xT, -zT, -yT
1H 3524.63 749.613 4.706 2048 1024
13C 7728.67 188.497 41 206 (acqu3s) 103
15N 8962.96 75.966 118 4 (acqu2s) 2
HNCOCA @ 750MHz
From acqus
NUC1/2/3 O1/2/3 SFO1/2/3 Center (O/SFO, ppm) TD (=-xN, -zN, -yN) -xT, -zT, -yT
1H 3524.63 749.613 4.702 2048 1024
13C 9989.95 188.499 53 206 (acqu3s) 103
15N 8962.96 75.966 118 4 (acqu2s) 2
HNCA @ 750MHz (same as its HNCOCA pair)
From acqus
NUC1/2/3 O1/2/3 SFO1/2/3 Center (O/SFO, ppm) TD (=-xN, -zN, -yN) -xT, -zT, -yT
1H 3524.63 749.613 4.702 2048 1024
13C 9989.95 188.499 53 206 (acqu3s) 103
15N 8962.96 75.966 118 4 (acqu2s) 2
The PoissonGap setup for a triple resonance experiment.
The Experimental Tab provides an estimation of T2 (J-coupling) so you can achieve maximum S/N at 1.2*T2 (factor term in poissongap.jar) .It just determines the total number of points required to reach 1.2*T2 as maximum evolution time (acquisition time).
Usually the evolution time is for an antiphase term in triple resonance experiments. However due to assumptions, setting antiphase/in-phase is not critical.
INPUT OUTPUT
1st ID (slow) 2nd ID (fast) vclist (t1) vplist (t2) FID total point matrix ~10% FID total point matrix
CBCACONH, HNCOCA, HNCA
F1(H) -> F2 (Calip. slow t1 -> Ca) -> F2 (C=O) -> F3 (N, fast t2) -> F1 (H, t3) - CBCACONH
F1(H) -> F3 (N) -> F2 (C=O) -> F2 (Ca, slow t1) -> F2 (C=O) -> F3 (N, fast t2) -> F1 (H, t3) - HNCOCA
F1(H) -> F3 (N) -> F2 (Ca, slow t1) -> F3 (N, fast t2) -> F1 (H, t3) - HNCA
C (61) N (32) 0 to 60 0 to 31 61x32 = 1952 206
HCCCONHgp3d3
F1 (H) -> F2 (Caliph, slow t1 -> Ca) -> F2 (C=O) -> F3 (N, fast t2) -> F1 (H, t3)
C (100) N (40) 0 to 99 0 to 39 100x40 = 4000 400
HCCCONHgp3d3
F1 (H, slow t1) -> F2 (Caliph -> Ca) -> F2 (C=O) -> F3 (N, fast t2) -> F1 (H, t3)
H (100) N (40) 0 to 99 0 to 39 100x40 = 4000 400
(1) Save output as expt.2d and randomized output as no
(2) Extract first column as t1_vclist points 13C
cat expt.2d | awk {'print $1'} > t1_vclist
(3) Extract second column as t2_vplist (randomized) points 15N
cat expt.2d | awk {'print $2'} > t2_vplist
(4) switch columns by pasting t1_vplist t2_vclist > sched.2d
More to come...
Relaxation Delay Setup
The relaxation delay schedule vdlist (variable delay list, default unit: sec) for T1 uses the equation based on the longest delay time (pulse length) collected.
First, for determining the longest delay time, we collect 1D 1H-NMR spectra in the same 2D relaxation experiment (set the indirect dimension -yN 1 and acquire (1H) direct dimension e.g. -xN 2048) at increasing delay times till the max. peak intensity decays to 20-30%.
For example in the vdlist, the longest T1 delay for 15N or 13C is 0.045s. For a 9 delay-time series, the T1 delay times are thus 0.001667, 0.
004714, 0.008660, 0.013333, 0.018634, 0.024495, 0.030867, 0.037712 and 0.045 based on ni=1...9 and nmax =9 and delay max = 0.045.
The final vdlist is scrambled into the order 1 (0.001667), 9, 2, 8, 3, 7, 6, 4, 5, 2, 4, 8 where the duplicated orders are 2, 4, 8. For 13C T1rho, the longest spin-lock time in the vplist (variable pulse list, default unit: us) e.g. 25000us (25ms) will be treated as the delay max.
Also, for example in the vclist (variable counter list, default: unitless integer) of 15N T2, the longest counter is 9, the final vclist is scrambled into 1, 9, 2, 8, 3, 7, 6, 4, 5 and repeats of 2, 4, 8.
Lastly, the vdlist or vclist is copied to the vdlist/ vclist variable in Topspin Bruker Program and the indirect (15N/13C) dimesion is set to the actual points (e.g. -yN 96 and -yT 48 for -yLAB 15N) to collect the series of 2D relaxations. The peak intensity decay for each residue is plotted against the delay times using an in-house python script with peak height uncertainties (or standard deviation based on duplicates) scaled by sqrt2 (Skelton et al. click).
References:
1. J Biomol NMR 2012 52:315
2. J Magn Reson, Ser B 1993 102:253