A new method for the comparison of 1H NMR predictors based on tree-similarity of spectra

Abstract

A methodology based on spectral similarity is presented that allows to compare NMR predictors without the recourse to assigned experimental spectra, thereby making the task of benchmarking NMR predictors less tedious, faster, and less prone to human error. This approach was used to compare four popular NMR predictors using a dataset of 1000 molecules and their corresponding experimental spectra. The results found were consistent with those obtained by directly comparing deviations between predicted and experimental shifts.

Fully automatic assignment of small molecules' NMR spectra without relying on chemical shift predictions

“…B, with δ = 0.5 ppm. This value was chosen to account for 1 H chemical shift fluctuations induced by variation in experimental conditions and for prediction accuracies of standard NMR predictors …”

“…When chemical shifts are taken into account, vector v consists of predicted chemical shift values v i , that may be obtained through any state‐of‐the‐art chemical shift predictor, e.g . for protons and for 13 C. Now, as noted earlier, although a typical predictor yields a single chemical shift value, one should acknowledge that there is a significant degree of uncertainty related to the quality of the prediction and to variability in the experimental conditions when working with 1 H NMR data and choose a similarity function that accounts for it. This can be carried out by associating the ‘exact chemical shift prediction’ with a ‘diffuse prediction’, which would be represented by a normal distribution around the predicted value, such as the one shown in Fig.…”

“…( A ) A normal distribution around a predicted 1 H chemical shift value v i represents a ‘diffuse’ prediction that accounts for the uncertainty in the result produced by the NMR prediction software and different experimental conditions . The height of the curve is related to the expectation of observing the peak at the corresponding chemical shift coordinate.…”

Complex mixtures by NMR and complex NMR for mixtures: experimental and publication challenges

“…[39] Although all these concerns may look like mere technicalities in programming scripts, they are in fact of central importance. In this sense, a recent evaluation of various NMR chemical shifts predictors [40] is quite revealing: it was found that, unsurprisingly, the best predictor was the one supported by the most comprehensive and well-curated database. The actual prediction methods used were not even of relevance, for no amount of ingenuity could overcome this 'merely technical' advantage.…”

Molecular search by NMR spectrum based on evaluation of matching between spectrum and molecule

“…The spectrum of the candidate molecule can be experimentally measured using NMR spectroscopy or simulated via quantum chemical calculation 10 . Various spectral similarity measures then become available for use 11 – 13 , which directly operate on raw spectra without an explicit annotation of the chemical shifts. This approach requires securing the spectra of all candidate molecules, which is difficult in practice.…”