State Key Laboratory of Magnetic Resonance and Atomic and Molecular Physics, Wuhan Institute of Physics and Mathematics, the Chinese Academy of Sciences, Wuhan 430071, China
2.
College of Chemistry, Central China Normal University, Wuhan 430079, China
3.
Department of Biological Sciences, National University of Singapore, Singapore, 117543
Received Date:
01 September 2017 Accepted Date:
02 November 2017 Available Online:
20 December 2017
Fund Project:
This work was supported by the National Key R & D Program of China (Nos. 2017YFA0505400, 2016YFA051201), the National Natural Science Foundation of China (Nos. 21735007, 21475146, 81227902, 21675170)
Abstract:
Sampling truncation, i.e. sampling ends before signal decays to zero, introduces wiggle-like artifacts that impair the quality of nuclear magnetic resonance (NMR) spectra. In multidimensional (mD) NMR experiments, the first few hundred or less data points are commonly sampled in the indirect dimension, and the truncation is unavoidable. Apodization can suppress truncation artifacts with cost of line-broadening. Linear prediction is also beneficial to the reduction of truncation artifacts. The worst situation is in the constant time (CT) type experiments, where the signal doesn't decay at all in the CT evolution dimension. In this contribution we proposed that, although iterative soft thresholding (IST) was rarely used due to the difficult parameter tuning, it was particularly suitable to suppress the truncation artifacts in the CT type NMR experiments. The simulation and experiments were performed to show the performance of this method. And the processing result was compared with a method proposed recently.
Figure 1.
Simulation showing the principle of truncation artifacts suppression using iterative soft threshold-donoho (IST-D). (A) a peak with truncation artifacts. (B) the extracted part of (A). (C) the truncation artifacts contribution of (B) plus itself. (D) the difference of (A) and (C). Clearly, the artifacts in (D) is weaker than those in (A). The subtracted part in this simulation is much larger than in real processing, for sake of clear demonstration
Figure 3.
Simulation of IST-D processing on truncated data. As the reference, spectrum (A) are obtained from the long data set (1024 complex points) by fast Fourier transform (FFT). Spectrum (B) is obtained from the short data set (180 complex points) by FFT. From the same data set, (C) is obtained by apodization and FFT, and (D) is obtained by IST-D
Figure 4.J-modulated [1H-13C] CT-HSQC spectra of 13C-15N labeled ubiquitin. The spectra shown in (A), (B) and (C) are from the same sampled data which is truncated in F1 dimension (13C), and share the same processing parameters in F2 dimension. In F1 dimension, a squared cosine window function is applied to produce (A), linear prediction and the same window function are applied on (B), and IST-D is performed on (C). The spectrum with much longer sampling time in F1 dimension is shown in (D) as reference. Only a small region of methyl groups are shown for sake of detailed display
Figure 5.
2D contour and 1D slices of J-modulated [1H-13C] CT-HSQC spectra of 13C-15N labeled ubiquitin. A reference spectrum shown in (A) was from the full sampled data, and it is same as Fig. 4D. From the same truncated sampled data, Fig.B and C are obtained using IST-D and SMILE, respectively. B is identical to Fig. 4C. Only a small region of methyl groups are shown for sake of detailed display. D－F are 1D slices along the dotted lines in A－C, respectively
Bostock M J, Holland D J, Nietlispach D. J. Biomol. NMR, 2012, 54(1):15-32 doi: 10.1007/s10858-012-9643-4
[19]
Jiang B, Luo F, Ding Y M, Sun P, Zhang X, Jiang L, Li C G, Mao X A, Yang D W, Tang C, Liu M L. Anal. Chem., 2013, 85(4):2523-2528 doi: 10.1021/ac303726p
[20]
Högbom J. Astron. Astrophys. Suppl. Ser., 1974, 15:417
Figure 1
Simulation showing the principle of truncation artifacts suppression using iterative soft threshold-donoho (IST-D). (A) a peak with truncation artifacts. (B) the extracted part of (A). (C) the truncation artifacts contribution of (B) plus itself. (D) the difference of (A) and (C). Clearly, the artifacts in (D) is weaker than those in (A). The subtracted part in this simulation is much larger than in real processing, for sake of clear demonstration
Figure 3
Simulation of IST-D processing on truncated data. As the reference, spectrum (A) are obtained from the long data set (1024 complex points) by fast Fourier transform (FFT). Spectrum (B) is obtained from the short data set (180 complex points) by FFT. From the same data set, (C) is obtained by apodization and FFT, and (D) is obtained by IST-D
Figure 4J-modulated [1H-13C] CT-HSQC spectra of 13C-15N labeled ubiquitin. The spectra shown in (A), (B) and (C) are from the same sampled data which is truncated in F1 dimension (13C), and share the same processing parameters in F2 dimension. In F1 dimension, a squared cosine window function is applied to produce (A), linear prediction and the same window function are applied on (B), and IST-D is performed on (C). The spectrum with much longer sampling time in F1 dimension is shown in (D) as reference. Only a small region of methyl groups are shown for sake of detailed display
Figure 5
2D contour and 1D slices of J-modulated [1H-13C] CT-HSQC spectra of 13C-15N labeled ubiquitin. A reference spectrum shown in (A) was from the full sampled data, and it is same as Fig. 4D. From the same truncated sampled data, Fig.B and C are obtained using IST-D and SMILE, respectively. B is identical to Fig. 4C. Only a small region of methyl groups are shown for sake of detailed display. D－F are 1D slices along the dotted lines in A－C, respectively
下载:
下载:
下载:
