June 13-15, 2018
Hotel Korpilampi, Espoo, Finland
Generating Unbiased Structural Alternatives for Automated Structure Verification
S. Golotvin, R. Pol, M. Elyashberg, D. Argyropoulos, and K. Kassam Tuesday, June 14th, 11:30–12:00 CET
Abstract: Automated structure verification (ASV) using NMR data is gaining acceptance as a routine application for qualitative evaluation of large compound libraries produced by synthetic chemistry. The simplest version of this confirms whether a proposed structure is consistent if it fulfils certain conditions. These are usually from 1D–1H NMR data  a combination of 1D–1H and 1H–13C HSQC spectra, or other 1D and 2D data.
Although it is easy to see when a proposed structure does not pass this "NMR filter", it is not trivial to guarantee that a proposed structure is correct. This is because the process of verifying structures against an NMR data set is inherently biased. A proposed structure is selected based on the expected knowledge and chemistry of the sample, and no alternative structures are presented as potential "better fits".
There are several ways to improve this method. One way is to use as many NMR experiments as possible for structure validation (1D–13C, 1H–13C HMBC, etc.). However, this can be time-consuming and expensive. The less costly method  is to generate and simultaneously verify several isomeric structures against the proposed one. This approach can highlight structural differences by gradually tightening NMR prediction tolerances until only a single structure remains.
While this method provides more confidence in the selected structure, it is still 'biased' since it only checks a few isomers, and may not include every isomer. As well, the alternative structures are 'biased' by the initially proposed structure since they are selected by an automatic algorithm or a human expert. As a result, other less similar structures that could be a better fit to the NMR data are not considered.
Here we present a method for generating alternative 'unbiased' structures for ASV, based on the structure generator  used in a CASE system. Using this approach we are able to generate every isomer that fits a particular NMR dataset without using any prior available knowledge. Practical aspects of this 'unbiased' verification are discussed and several examples are shown.
 S. Golotvin, E. Vodopianov, R. Pol, B. Lefebvre, A. Williams, R. Rutkowske, T. Spitzer, Mag. Res. Chem. 45(10), 803-813, (2007).
 S. Golotvin, R. Pol, R. Sasaki, A. Nikitina, P. Keyes, Mag. Res. Chem. 50(6), 429-435, (2012).
 M. Elyashberg, A. Williams, K. Blinov, "Contemporary computer-assisted approaches to molecular structure elucidation", RSC, Cambridge, (2012).