Biomedical Engineering Reference
In-Depth Information
can also design a pulse scheme with C
ali
and C9 selective pulses and with
optimal delays for C
a
N and C9N transfers. This experiment maps all CN
correlations in a protein in a single experiment. This may not be possible in
uniformly
13
C-labelled samples as the evolution of
1
J
CaC9
within a pulse
sequence as well as in a detection period is inevitable. Broadband excitation
may be a challenge in high-magnetic fields, however, recent advances in
broadband pulse design can cope with this problem.
64,69
While the NCA 2D spectrum shown here provided sufficient resolution
(Figure 2.3), one can also apply non-uniform sampling to improve resolution
without extending the overall measuring time. For this purpose, a series of
linearly and non-linearly sampled 2D
13
C
a
-detected NCA experiments were
recorded on an alternately
13
C-labelled B1 domain of protein G (GB1) and
analyzed with respect to the effect on spectral quality, S/N, recovery of very
weak peaks, and fidelity of peak positions.
70
The results are summarised in
Figure 2.5. A total of five NCA experiments (labelled A1-A5) were recorded.
Two NCA reference spectra of 256 uniformly sampled increments in which the
number of scans per increment was (A1) 2 and (A5) 8 were measured in 22 min
and 1.5 h, respectively. We then recorded three spectra with one quarter of the
increments selected but accumulating eight scans per increment. These
required 22 min of measuring time each and were sampled as follows: (A2)
64 first of 256 uniformly sampled increments, (A3) 64 out of 256 increments
randomly selected with uniform sampling density, and (A4) 64 out of 256
selected by sine-weighted Poisson-gap sampling (SPS). The non-uniformly
sampled spectra were subsequently FM-reconstructed. The spectrum A2 that
sampled the 64 first increments was extended to 256 points using linear
prediction with an order parameter of 30.
A SPS scheme combined with forward maximum-entropy (FM) reconstruc-
tion demonstrated the superior performance over other methods in the analysis
includes all resonances [Figure 2.5(B)] as well as only small peaks
[Figure 2.5(C)]. The relative signal amplitudes are well preserved (high R
2
values). We also find that S/N is enhanced up to 4-fold per unit of data
acquisition time relative to traditional linear sampling. As previously reported,
linear prediction can cause small changes in peak positions, and this is clearly
seen for the peak at the nitrogen position of 115.1 ppm (top dotted line in
Figure 2.5). Such chemical shift changes are not observed in the SPS/FM
reconstruction approach. Furthermore, linear prediction clearly suffers from
significantly lower resolution, which matters for crowded spectral regions.
It is also interesting to compare how NCA and NCO peak intensities
decrease as a function of the molecular weight. Figure 2.6(A) shows a
simulation using the relaxation parameters estimated in Figure 2.1. In
Figure 2.6(B), the 1D NCA and NCO spectra are shown recorded at 10 K
as well as 90 K conditions. The advantage of the
13
C
a
detection is obvious from
the simulation especially for a high magnetic field. One can also see a clear
difference even in the spectra recorded at lower magnetic field (11.7 T). The
NCO signals are mostly absent at 90 K conditions except for signals from
