Automatic NMR Solvent Recognition wizard:
an example of an Artificial Intelligence at work

by Stanislav Sykora (Extra Byte, Italy), and
Carlos Cobas, Felipe Seoane, Pablo Monje, Esther Vaz, et al (Mestrelab Research, Spain)

delivered at the Mestrelab User Meeting, SMASH 2010, Portland, OR (USA), September 25, 2010.

Presentation SLIDES ( PDF, 1239 kB) NMR Articles and Posters
Please, e-mail me your comments and questions. I will collect them verbatim and answer the questions on this page.
Click here to view all talks presented at this User meeting of Mestrelab Research.

Presentation Abstract

The recognition of solvent peaks is part of what we call automatic peaks editing. In a sense it is a refinement of the GSD (Global Spectral Deconvolution) process which ends up with a table of peaks present in a 1D spectrum. The AutoEdit procedure than attempts a classification of the peaks according to a number of criteria (compound, reference, solvent, impurity, 13C satellite, labile, weak, etc.). The classification is an essential step required before attempting any structure verification, be it automatic (ASV) or manual.

The solvent recognition, in particular, is definitely non-trivial. At some point, the development of this procedure started implying that the system must be "teached" how to apply all the criteria and tricks which usually enable an experienced chemist to decide which peaks in the spectrum belong to the primary and secondary solvent. The result is a sophisticated and ever more capable NMR solvent-recognition wizard, based on modern scoring techniques and a long list of scoring criteria to be applied to all candidate peaks.

In this presentation, the criteria are discussed and the wizard's performance is illustrated.

Stan Sykora's Publications, Posters and Courses & Talks Stan's Library | Stan's HUB
Copyright ©2010 Sýkora S.,    DOI: 10.3247/SL3Nmr10.007 Page design by Stan Sýkora