Benchtop NMRs Challenge Data Analysis

The interest in high-resolution bench-top NMR systems is very topical, as institutions consider budgets, TCO, new uses for NMR, versatile deployment options, and a cost-effective new teaching tool. Of course, many of us of a “certain age” will have fond memories of significant research performed on instruments operating at ≤100MHz for 1H. The wave of new systems has great appeal, and stems from advances in permanent magnet designs that use rare earth metals. Interest in this technology runs the spectrum from MRI scanners to the bench-top systems we will discuss today. 

Bench-top systems can currently be used for TD-NMR, MRI, and high-resolution NMR. This blog focuses on software concerns high-resolution spectra and issues that come into sharper focus with these systems. Are the existing processing packages and algorithms adequate? To begin to try and answer this question let’s consider the possible uses as we see for these systems today, and anticipate the new challenges for each of these.

Better put, perhaps, is “forewarned is forearmed!” We know that compromises must be faced in terms of lower dispersion at lower field, but is the problem significant? You will likely have a high-resolution NMR spectrum of your compound(s) of interest, but knowing the issues at low-field can be achieved before you put on your lab coat or spend a dime? Figure 1 shows a recorded spectrum of a marketed drug, pioglitazone, and the spectrum you could expect at 80 MHz – ignoring S:N issues. So, when we have the hi-field spectrum available, we can quickly see and anticipate the extent of inherent problems associated with lower-field spectra.

Recorded spectrum of a marketed drug, pioglitazone

So much of NMR analysis comes down to accurate peaks detection and determining their properties. The process is more difficult when there is significant overlap, S:N is restrictive, and dynamic range is an issue – all issues mentioned in the table. One, efficient solution to the problem is line deconvolution. With conventional algorithms the computational burden scales to n4, where n is the number of lines. This approach becomes impractical when n exceeds about 10. The answer is to use a different algorithm, such as GSD (Mestrelab Research). With algorithms if this type, an entire spectrum having 100s of lines can be quite accurately deconvolved in a few seconds. In Figure 2 [spectrum courtesy of Dr. Andrew McLachlan, Thermo Fisher Scientific] you see a 1H NMR spectrum of ethanol recorded at 45MHz. The water signal is intrusive, and S:N is 67 for the peak at 0.92 ppm. The spectrum is colored maroon, the GSD peaks individually are orange, summed they are blue, and the residual is colored cyan. We see that GSD does a very good job with the solute peaks despite some challenges that will be typical of low-field spectra.

1H NMR spectrum of ethanol recorded at 45MHz

Working with data arrays at low field invariably leads to a good number of challenges. The data in Figure 3 [courtesy of Dr Maria Silva Elipe, Amgen] are for a simple reaction, where the total methyl singlet resonance was monitored by recording NMR spectra at 45MHz. You might think that there is no hope in recovering useful information from these data, and yet the extracted curves look quite reasonable (top graph). We can be spoiled by the spectral quality obtained using high-field instruments: analogous data from low-field systems may not necessarily look as “pretty,” but can still be useful. Here we have not discussed the use of chemometric techniques – another very useful possibility with complex, overlapping data.

NMR spectra at 45MHz

Working with low-field, high resolution NMR systems will pose challenges to the data analysis capability. Even in these early days, we see that one can go quite some distance to meeting these and for the instruments to be productive. Even before special algorithms are developed, the future looks bright!

Back to February Tech Talk