SMASH – Small Molecule NMR Conference

Dimitris Argyropoulos, Sergey Golotvin and Uliana Bortnik

The synergy of modern NMR instruments, automated sample handling, and data interpretation through Automated Structure Verification (ASV) has ushered in an era of unparalleled high-throughput NMR capabilities. These automated systems are typically configured with standardized acquisition parameters¹ to cater to a diverse array of samples efficiently. Nevertheless, despite scientists’ best efforts to optimize these parameters, instances arise where these parameters yield spectra with peaks deviating from their anticipated positions or phase.

For instance, employing an HSQC experiment with an F1 width set to 0-160 ppm for samples containing aldehyde or Si­CH₃ groups may lead to peaks appearing at the periphery of the F1 window, if not entirely beyond it and aliased over². Additionally, modern adiabatic HSQC-DEPT pulse sequences, designed to enhance experiment sensitivity³, may yield inaccurate results in ASV systems for samples with cyclopropyl or acetylene protons. Peaks corresponding to these groups may manifest with the opposite phase in the resultant spectra.

Discrepancies between expected and observed peaks can introduce significant disruptions to high-throughput workflows, forcing chemists to manually review datasets and, in some cases, re-record spectra with parameters more suitable for that sample. However, by providing the ASV system prior knowledge of the proposed structure, and training it to recognize peaks with unexpected presentations, these disruptions can be avoided.

This poster examines such an automated solution that seamlessly addresses this challenge, demonstrating its effectiveness in high-throughput laboratories. A range of examples showcasing each case will be presented, underscoring the robustness and reliability of the proposed approach.

1. Golotvin, S.S., Vodopianov, E., Pol, R., Lefebvre, B.A., Williams, A.J., Rutkowske, R.D., Spitzer, T.D., Reson. Chem., 45, 2007, 803-813.
2. See for example Claridge, T., “High Resolution NMR Techniques in Organic Chemistry”, Elsevier, 3^rd Edition, 2016
3. Boyer, R. D., Johnson, R., Krishnamurthy, K., Mag. Res. 165, 2003, 253-259.

Alexander Bourque, Sergey Golotvin, Maxim Kisko, Rostislav Pol and Dimitris Argyropoulos

NMR data is invaluable in determining the structures of new and/or unknown compounds using Computer-Assisted Structure Elucidation (CASE). [1] Starting with a molecular formula, most existing CASE systems will solve the problem by defining a set of constraints derived from the observed chemical shifts (usually at least ¹H and ¹³C) and the analysis of the observed correlations in the various types of 2D spectra (HSQC, HMBC, COSY, and others). All possible chemically sensible structural isomers are generated using a mathematical procedure, including all possibilities by renumbering the atoms. The ones satisfying the previously implied constraints are selected to be ranked, usually by the deviation between the experimental and observed chemical shifts. Despite the huge advances in computing power over the past years, this structure generation step remains the bottleneck in CASE workflows. This becomes especially problematic as the overall number of atoms increases and heteroatoms are involved, which makes the computational task formidable.

To reduce the time required for this structure generation step, one must increase the number of constraints. This can be done either by recording additional spectra that would reveal more correlations (e.g., C-C correlation spectra like ADEQUATE and INADEQUATE) and/or identifying some known fragments of the structure using the existing data. While it is not always possible or feasible to record additional spectra, identifying known fragments can more easily be achieved using fragment libraries or the ability of NMR spectroscopists to recognize familiar spectral patterns, specific to particular fragments.

In this poster, we present an automated approach that mimics this process that the NMR expert would do and identifies spectral patterns characteristic of phenyl fragments based on the chemical shifts, observed connectivities, and symmetry. Depending on the resolution of the signals, this method would produce one or more Molecular Connectivity Diagrams (MCDs) that already include the phenyl group(s) thus, significantly reducing the complexity of the problem and shortening the elucidation time. We will compare the time taken to elucidate several structures with and without this approach to demonstrate its utility.

1. E. Elyashberg, A.J. Williams. “Computer-based Structure Elucidation from Spectral Data. The Art of Solving Problems”, Springer, Heidelberg, 2015, 454 p