Mestrelab Predictor

Mnova comes with a new module for the fast and accurate prediction of 1H and 13C NMR spectra. It consists of a novel proprietary algorithm based on a machine learning algorithm (Mestrelab Predictor) which has been training with a large database of experimental assigned 1H and 13C chemical shifts as well as proton scalar coupling constants.

More information about Mestrelab Predictor can be found here and in this paper.

Running a Prediction

Once a molecule is available in Mnova, predicting a 1H or 13C spectrum can be done by simple selecting that option either from the Predict menu or by right clicking on the molecule and selecting the prediction from corresponding contextual menu.

First time a prediction is run, it can take a few seconds for the mathematical prediction model to be loaded in memory. This is done only once, so subsequent predictions will be extremely fast, although the model will be unloaded from memory if no predictions are run during five minutes.

Prediction options

From/To: Lower and upper limits for the chemical shift scale

Number of Points: Number of points used to build the predicted spectrum. Actually, Mnova predicts a time domain signal (FID) which is automatically Fourier Transformed to yield the final predicted frequency domain spectrum.

Line Width: Line width at half height of all peaks in the predicted spectrum.

Solvent: Prediction solvent.