NMR Predictors

Many people want to learn whether a bench top NMR instrument is suitable for them. Here's an application chemist's evaluation process.

When people ask me about benchtop NMR instrumentation, invariably I'll ask about their workflow, their present NMR access and usage, and the typical sample types that they want to analyse by NMR. The prediction tools in Mnova™ are extremely helpful to me as I can show the sort of results they can expect from their samples at an early stage, before planning a site visit. This works well for 'small molecule' people; typically synthetic organic chemists working on drug like molecules with a molecular weight of around 500 Da.

Using a chemistry drawing package such as MarvinSketch, or Mnova's built in structure drawing tool, I can draw a chemical structure. This, once attached to a blank page can be used to generate a predicted spectrum at a frequency of my choosing e.g. 500MHz, or 82MHz. Samples with lots of peaks will exhibit more overlap at relatively low field, and visualising this is helpful when attempting to manage the expectations of someone who is used to data from high field instruments. The picture below shows spectra of ethyl acetate at 45MHz, 82MHz and 400MHz, and these were generated with Mnova and the Modgraph NMR Predict module.

When performing instrument demos, I often begin by injecting something simple e.g. ethyl acetate, and then acquiring a 1H spectrum. Reviewing data from the little NMR spectrometer in the Mnova software, and performing a handful of operations (referencing, peak picking, integration) produces a processed spectrum, with a nice triplet and quartet splitting plain to see. For the extra wow, I overlay a predicted spectrum so people can marvel at the comparative results. So many people are not familiar with NMR prediction software, and people's curiosity is often then piqued…. It is not such a huge logical inference for someone to then ask: "Can the software elucidate the structure from the spectrum?" What a question! I used to work for ACD/Labs, and I've had some exposure to the various computational approaches to doing this, having played with their Structure Elucidator Suite, and from hearing talks at the SMASH conference. Knowing the types and quality of data needed to attempt this for small molecules (generally 13C, HSQC, HMBC and molecular weight) this is not a trivial task for a software package. With tools such as Structure Elucidator, the answer is maybe, in that "potential structures that are consistent with the data provided" can be generated. Much has been written, discussed and developed over the years to enable this to be possible, and I'm not going to go into this here. However, I will remark that I have found discussions on this topic with my former ACD/Labs colleagues Patrick Wheeler and Ryan Sasaki, and GSK's John Hollerton fascinating. Anyone used to elucidating structures will no doubt scoff at the idea of using a bench top NMR instrument for such purposes as they do not provide high enough resolution. The point I want to make is that with the availability of bench top NMR instrumentation, and their use in non-traditional environments, many of the tools used by spectroscopists will start to become familiar and come to the attention of non NMR experts for use in niche applications. Now, back to NMR prediction software; this enables a spectrum to be generated from a chemical structure. This is enormously helpful as an aid to interpreting experimental data and is generally accurate, but can be prone to error if e.g. the query structure is not well represented by the records in the software's predictor database (HOSE), or if measurements are performed at different concentration, pH, temperature or in different solvents. Therefore, one must exercise caution about the quality of predicted spectrum vs. real experimental data.

Back to April 2014 Tech Talk