NUS for ASAP-HSQC experiments

Non Uniform Sampling (NUS) speeds the collection of multidimensional NMR spectra by measuring only a fraction of the data and predicting what was omitted. The quality of the reconstructed data depends on many factors and what works well for one experiment many not work well for others. The last two posts examined how NUS affects COSY and HMBC experiments. This post examines if NUS impacts the ASAP-HSQC experiment more than the traditional HSQC.

ASAP (Acceleration by Sharing of Adjacent Polarisation) is an innovation that reduces NMR data collection time by allowing the time between scans to be dramatically reduced. For any NMR experiment more complicated than a simple 90^o pulse and record, reducing the time between scans generates many artifacts as the magnetisation does not have time to return to equilibrium and the carefully planned manipulation of the magnetisation is disrupted. ASAP actively returns magnetisation to equilibrium using a short TOCSY mixing period in the relaxation delay to transfer magnetisation from protons attached to spin inactive nuclei, such as ¹²C in a ¹³C experiment. Thus, ASAP is useful in heteronuclear experiments on unlabelled samples where the majority of the sample does not produce a signal.

As well as reducing the time between scans, the ASAP-HSQC parameters implemented at the SSPPS Facility reduce the time spent recording the signal. This speeds the experiment further, and also reduces the fraction of time spent decoupling. If decoupling is active for too high a fraction of the experimental time then the probe can be damaged. The ASAP-HSQC parameters in use at the Facility do not cause excessive probe heating and appear to be safe. They also reduce the experimental time by a factor of six.

To see if NUS impacts the ASAP-HSQC more than the standard HSQC the halogenated steroid sample from the previous posts was used. Spectra were collected identically except for the amount of NUS sampling and the pulse sequence used. Processing with TopSpin 4.0 used cosine squared apodisation in both dimensions and four-fold linear prediction in the indirect dimension. In the ASAP-HSQC spectra four-fold linear prediction was also used in the directly detected dimension. Spectra were collected at four different levels of NUS and are shown below plotted with the same threshold and contour intervals.

Both sets of spectra show a loss of signal intensity as the amount of sampling is reduced, but this is to be expected as the measurement time is reduced. (The 12.5% ASAP-HSQC spectrum was recorded in less than a minute.) The ASAP-HSQC spectra are less intense than the standard HSQC, so the loss of intensity with reduced sampling is more noticeable. Noise does not appear to increase with reduced sampling as dramatically as with the HMBC and COSY experiments.

Examining slices along the ¹³C dimension shows that the experiments are remarkably robust. Slices from the HSQC spectra taken at 1.1 ppm in the ¹H dimension are shown below.

The peaks at 15, 24, 40 and 57 ppm are all true signals. It is only when the sampling is reduced to 12.5% that the smaller true signals become indistinguishable from noise and significant artifacts appear (74, 45 and 27 ppm). Matching slices from the ASAP-HSQC spectra are shown below.

The ASAP-HSQC slices show more baseline noise and artifacts appear earlier than in the standard HSQC, at 25% sampling, but all the true signals are present. For many users the six fold reduction in acquisition time may be worth accepting the increased artifacts, and most users are now opting for the ASAP-HSQC, but for those that prefer the traditional HSQC those parameters are still available. The default parameters have the ASAP-HSQC using 50% NUS, while the standard HSQC uses 25%. These values can easily be changed in the IconNMR interface.