PANIC, April 26-30, 2020 | ACD/Labs
ACD Labs Logo
MENU

PANIC

April 26-30, 2020
La Jolla, CA, USA



Poster Schedule

Detection of Multiplets in 13C Spectra Using Structure Aware Algorithms
Jessica Litman, Sergey Golotvin, Rostislav Pol, Vladimir Mikhailenko and Dimitris Argyropoulos
View Abstract

Detection of Multiplets in 13C Spectra Using Structure Aware Algorithms

Jessica Litman, Sergey Golotvin, Rostislav Pol, Vladimir Mikhailenko and Dimitris Argyropoulos

Advanced Chemistry Development, Toronto, ON, Canada

Software encapable of Automatic interpretation of NMR spectra are highly demanded in the modern chemical and pharmaceutical industry.  Every day scientists record hundreds of NMR spectra that need to be interpreted.  Spectra interpretation can be quite daunting and prone to human error, which makes it a good candidate for automation.  The first step in an automated interpretation of NMR spectra is the reliable detection of the peaks and multiplets in the recorded spectra.  This seemingly simple task can become quite complicated if the structures contain heteroatoms such as 19F and 31P that give rise to additional peak splittings in the 1D 13C and heteronuclear 2D spectra.  With up to 30% of the commercial pharmaceutical APIs and agrochemicals containing either a 19F and/or a 31P atom, the need for a reliable peak detection in these chemicals is quite clear.

1H decoupled 13C spectrum of bis(4-fluorophenyl)(oxo)phenyl-(gamma)-phosphane, showing the peaks and extensive couplings observed for a structure with only 8 magnetically equivalent carbons
Figure 1: 1H decoupled 13C spectrum of bis(4-fluorophenyl)(oxo)phenyl-λ5-phosphane, showing the peaks and extensive couplings observed for a structure with only 8 magnetically equivalent carbons

The traditional way of accomplishing this task is to select all the visible peaks in the 1D 13C spectrum and define them as singlets.  For specific patterns, for example three evenly spaced peaks with a 1:2:1 intensity ratio, a triplet will be defined and treated as a single multiplet.  A similar procedure is followed for patterns that fit into quartets, however, there is no reliable method for doublets.  Moreover, the whole procedure fails when the two outer peaks of the multiplet with lower intensities are not observed at all.   Another common problem is the overlapping of multiplets, since these are usually quite broad and the 19F and 31P nuclei are usually coupled to more than one carbon.  In this latter case, the procedure will fail as well.  Peak interpretations can become even more complicated in 2D spectra with characteristic diagonal peak patterns causing confusion even to some experts.  Even though for F-containing chemicals the problem can be alleviated by using costly required hardware and recording a spectrum with additional 19F decoupling, there is no such option for 31P.

Here we present a more generalized method for multiplet detection in such peaks, which resembles the way a human analyst would follow.  Since a proposed structure usually exists with the positions of the fluorine and phosphorus atoms defined, the human expert would specifically know what to look for, instead of starting from zero.  The way this can be implemented in software is by predicting the spectrum of the proposed structure and then looking for similar patterns in the experimental spectra.  If such patterns are found then the multiplets are defined correctly and the analysis can proceed without getting a false negative result, and need for manual inspection. This study will demonstrate examples of overlapped and low-intensity multiplets arising from 19F and/or 31P couplings and how these are consistently detected using this approach, as well as characteristic 2D spectra with the associated problems completely resolved.

Efficient Approaches for Addressing Spectral Ambiguities in Computer Assisted Structure Elucidation (CASE) Systems
Dimitris Argyropoulos, Rostislav Pol, Mikhail Elyashberg, and Sergey Golotvin
View Abstract

Efficient Approaches for Addressing Spectral Ambiguities in Computer Assisted Structure Elucidation (CASE) Systems

Dimitris Argyropoulos, Rostislav Pol, Mikhail Elyashberg, and Sergey Golotvin

Advanced Chemistry Development, Toronto, ON, Canada

Since their development over 50 years ago, Computer Assisted Structure Elucidation (CASE) systems (or Expert Systems, ES) have significantly facilitated the de novo structure elucidation of both natural and synthesized organic compounds, especially in cases where using the traditional (manual) methods would has been very challenging or even impossible to perform [1,2].  Current ES are based on 1D and 2D NMR spectra, given that the molecular formula has already been determined by HR-MS.  At present, there are several free and commercially available CASE systems offering the following main advantages [3]: i) ES deliver all (without any exception) structures which can be deduced from a given set of NMR data; ii) Application of fast empirical methods for NMR chemical shift prediction allows the program to select the most probable structure; iii) If necessary, DFT based chemical shift calculations are used to confirm the selected structure; iiii) ES are now capable of suggesting a 3D model of the elucidated structure.

Despite all the developments, ES are still susceptible to a series of limitations which impede structure elucidation by a human expert.  These limitations are mainly associated with the ambiguity of the experimental data, as well as the overlapping characteristic chemical shift values in NMR spectra.  Experimental ambiguity can result from low resolution 2D spectra, which prohibits the confident assignment of the closely spaced signals.  Further, ambiguity can often be due to the hybridization state of carbon nuclei that appear in the same regions of NMR spectra.  For example, a 13C NMR signal observed at 90 ppm can belong either to a sp2 or a sp3-hybridized carbon connected to one or two oxygen atoms.  If a molecule contains atoms that can have variable valences (e.g., N and/or P), all their possible valences should be explored during structure generation.  Ambiguity can also be observed in some cases where carbon nuclei show no HMBC or COSY correlations; this is especially evident in molecules with hydrogen deficiency which consequently appear as "floating", i.e., potentially could be connected to any other atom.  The presence of "floating" atoms significantly increases both the size of the output and the time of structure generation.

In order to remove uncertainty from spectroscopic data, additional experiments are usually carried out; for example high resolution 2D NMR spectra, such as highly inflated NUS or band selective spectra are collected to reliably assign the correlations.  Additional spectroscopic data (IR, Raman, UV-Vis) could help to identify the characteristic groups and resolve the hybridization or valence status.  However, such solutions are not useful in all cases.

In such cases, the only remaining solution is the exhaustive investigation of all alternatives ensuing from the presence of any ambiguity.  For example, if there are five carbon nuclei with signals in the range of 70-120 ppm, then 25=32 combinations of hybridization (sp3 or sp2) have to be generated and checked. This could significantly extend the structure generation time.

A typical Molecular Connectivity Diagram from a CASE ES:  Dotted lines represent ambiguous correlations, cyan carbon atoms are of either sp3 or sp2 hybridization, several
Figure 1: A typical Molecular Connectivity Diagram from a CASE ES:  Dotted lines represent ambiguous correlations, cyan carbon atoms are of either sp3 or sp2 hybridization, several "floating" atoms are visible.

In this poster by taking advantage of recent programming developments, we will present approaches that enable structure elucidation, under conditions where the initial data contains many ambiguous assumptions.  Examples will be presented, and the strengths together with the limitations of each approach will be discussed.

[1] Blinov, K.A.; Elyashberg, M.E.; Martirosian, E.R.; Molodtsov, S.G.; Williams, A. J.; Sharaf, M. M. H.; Schiff, P. L. Jr.; Crouch, R.C.; Martin, G. E.; Hadden, C.E.; Guido J.E.; Mills, K.A., Quindolinocryptotackieine: the Elucidation of a Novel Indoloquinoline Alkaloid Structure Through the Use of Computer-Assisted Structure Elucidation and 2D NMR.  Magn. Reson. Chem., 41, 577-584 (2003)

[2] M.E. Elyashberg, K.A. Blinov, S.G. Molodtsov, A.J. Williams., Elucidating "Undecipherable" Chemical Structures Using Computer Assisted Structure Elucidation Approaches, Magn. Reson. Chem., 2012, 50, 22-27

[3] M. Elyashberg, D. Argyropoulos. NMR-based Computer-assisted Structure Elucidation (CASE) of Small Organic Molecules in Solution: Recent Advances. eMagRes, 2019, Vol 8: 1–16. DOI 10.1002/9780470034590.emrstm1618.