ACD Labs Logo
MENU

SMASH

August 30-September 2, 2021
Online

Presentation

Session: Let In Silico do the Hard Work—Advances in Computational NMR
Evaluation of the Benefit and Informing Capability of 2D NMR Experiments for Structure Elucidation Using CASE Software
Dimitris Argyropoulos, NMR Business Manager, ACD/Labs
THURSDAY, SEP., 2ND, 12:20–12:35 PM

View Abstract

Evaluation of the Benefit and Informing Capability of 2D NMR Experiments for Structure Elucidation Using CASE Software
Dimitris Argyropoulos, NMR Business Manager, ACD/Labs

Computer Assisted Structure Elucidation (CASE) has been around for more than 50 years [1,2]. It has experienced a significant boost after the introduction of routine 2D NMR experiments in the 1990's, as the increased information content offered allowed much more complex problems to be addressed. The evolution of computers and the processing power of modern CPUs allowed CASE to be a tool of reference for resolving the unique and unprecedented structures of natural products [3]. Some common questions encountered when spectroscopists are introduced to CASE are regarding the minimum set of experiments required to solve a problem by CASE and what is the value and informing capability of advanced NMR experiments. Although the answers vary depending on the problem complexity, in this poster we will try to tackle both questions, by demonstrating a few cases.

We will look at three example structures to answer these questions more efficiently: 2-Ethylindanone, Spirodactylone [4] and a xanthone-class natural product [5-6]. In the first example, which is a very simple molecule, we will examine the benefits of basic NMR experiments and how they affect the calculation time. The second example is a challenging natural product, and we will examine the benefits and potential problems that are encountered when using modern experiments like LR-HSQMBC, in addition to the traditional HSQC and HMBC. Here, we will also study the influence of manually adjusting atom properties, such as hybridization and connection to heteroatoms. The final example serves for considering the inter-relation between HMBC and INADEQUATE when a CASE system is used.

Although a simple structure such as 2-Ethylindanone can be solved using a modern CASE system with only a 1D 13C and a 2D HSQC spectrum, adding a 1D 1H spectrum will result in a factor of 4 reduction in calculation time. This simple structure can be also solved with only COSY and HSQC spectra, without any 1D spectra at all. The CASE system can explore all options for the quaternary carbons that are not observed. However, using an HMBC instead of a COSY reduces the calculation time by a factor of 60.

Similarly, for Spirodactylone, a proton deficient molecule, adding appropriately defined long range correlations observed in LR-HSQMBC, that are not visible in HMBC, reduces the calculation time by a factor of almost 2000. Similar improvements are observed when specific information about carbon atoms that belong to carbonyl groups or connected to heteroatoms, are included. These are a few key points that the expert spectroscopist would consider when solving the structure manually and would not by any means bias the result.

In the next example, we study the alternatives to using the INADEQUATE experiment. INADEQUATE is considered to be the holy grail of NMR experiments as it can be invaluable for elucidating the structures of proton deficient molecules. However, this technique suffers from very low sensitivity and requires either large sample amounts or significant instrument time, usually days. Since a CASE system can explore all connectivity possibilities much faster compared to a human, the question is whether one could omit the INADEQUATE and instead work with a highly sensitive HMBC. For the case of the xanthone-class natural product studied here, a CASE system can give the correct answer in ca. 50 minutes, with some obvious edits to the Molecular Connectivity Diagram. This is a significantly less amount of time than what is required to record an INADEQUATE.

Exact details of the type of spectra used, and the calculations performed will be shown, as well as comparative tables on the achieved improvements in performance.

  1. Elyashberg, M.E., Argyropoulos, D. (2019). NMR-Based Computer-Assisted Structure Elucidation (CASE) of Small Organic Molecules in Solution: Recent Advances. eMagRes., 8(3), 239–254.
  2. Elyashberg, M., Argyropoulos, D. (2021). Computer Assisted Structure Elucidation (CASE): Current and future perspectives. Mag. Reson. Chem., 59(7), 669–690.
  3. Elyashberg, M.E., Williams, A.J. (2015). Computer-based Structure Elucidation from Spectral Data. The Art of Solving Problems (p. 454). Springer-Verlag Berlin Heidelberg.
  4. Kang, U., Caldwell, D., et al. (2019). Elucidation of Spirodactylone, a Polycyclic Alkaloid from the Sponge Dactylia sp., and Nonenzymatic Generation from the Co-metabolite Denigrin B. Org. Lett., 21(12), 4750–4753.
  5. Boudesocque-Delaye, L., Agostinho, D., et al. (2015). Antibacterial Polyketide Heterodimers from Pyrenacantha kaurabassana Tubers. J. Nat. Prod., 78(4), 597–603.
  6. Omolo, J.J., Maharaj, V. et al. (2012). Bioassay-Guided Investigation of the Tanzanian Plant Pyrenacantha kaurabassana for Potential Anti-HIV-Active Compounds. J. Nat. Prod., 75(10), 1712–1716.

Poster

A Strategy for the Best Candidate Selection from an Ensemble of Plausible Structures
Alexander Waked, Dimitris Argyropoulos, Maxim Kisko, Mikhail Elyashberg and Sergey Golotvin

View Abstract

A Strategy for the Best Candidate Selection from an Ensemble of Plausible Structures
Alexander Waked, Dimitris Argyropoulos, Maxim Kisko, Mikhail Elyashberg and Sergey Golotvin

A key step of Computer Assisted Struccture Elucidation (CASE) [1,2] is the selection of the most plausible structure from the set of structures generated using the spectroscopic data. A similar challenge exists with Automated Structure Verification (ASV) [3] systems where more than one proposed structures are defined or where the strategy of Unbiased Verification has been employed.

The traditional way of solving this in a CASE system is to predict the shifts of the atoms in the structures, compare them with the experimental shifts observed, and calculate the average and maximum deviations, usually for 1H and 13C. In ASV, a Match Factor, MF, is usually calculated using a different procedure. This MF is still based in the differences between calculated and experimental chemical shifts, but also takes into account the observed multiplicites, integrals, and observed correlations. The problem with both approaches is that, even though they can differentiate between correct and very wrong structures, it is not uncommon that multiple structures will have very similar scores, not allowing a safe conclusion about which is the most plausible one.

In recent years, the DP4 methodology has been suggested, for the determination of stereochemistry [4]. Like previous approaches, this methodology is based on predicting the chemical shifts, but it also takes into account the prediction error distribution.This is a multiplicatively calculated statistical, value that gives the probability of a particular structure from a set to be the correct one.

In this poster, we present the results we obtained when using DP4 style metrics to rank the results obtained in CASE and ASV runs. We studied a total of over 100 datasets, including both published and proprietary compounds. The datasets included at least 1D 1H and 13C spectra, as well as 2D HSQC and HMBC spectra. The 1H and 13C predicted chemical shifts were calculated using both the HOSE-codes based approach [5] and the proprietary Neural Networks. The actual prediction error distribution curves for each method were calculated using a set of 36,000 1H and 52,000 13C chemical shifts from 3,100 fully assigned structures that were not present in the predictor training dataset. It is shown that the DP4 method identified the correct structure in the vast majority of the cases. The actual results and a further discussion of the advantages and disadvantages of the method will be presented.

  1. Elyashberg, M.E., Blinov, K.A., Molodtsov, S. G., Williams, A. J., Martin, G. E. (2004). Structure Elucidator: A Versatile Expert System for Molecular Structure Elucidation from 1D and 2D NMR Data and Molecular Fragments. J. Chem. Inf. Comput. Sci., 44(3), 771–792.
  2. Elyashberg, M.E., Argyropoulos, D. (2019). NMR-Based Computer-Assisted Structure Elucidation (CASE) of Small Organic Molecules in Solution: Recent Advances. eMagRes., 8(3), 239–254.
  3. Golotvin, S.S., Pol, R., Sasaki, R.R., Nikitina, A., Keys, P. (2012). Concurrent combined verification: reducing false positives in automated NMR structure verification through the evaluation of multiple challenge control structures. Mag. Res. Chem., 50(6), 429–435.
  4. Smith, S., Goodman, J. (2010). Assigning Stereochemistry to Single Diastereoisomers by GIAO NMR Calculation: The DP4 Probability. J. Am. Chem. Soc., 132(37), 12946–12959.
  5. Bremser, W. (1978). Hose—a novel substructure code. Anal. Chim. Acta, 103(4), 355–365.