Prediction of a 1H NMR spectrum from the molecular structural formula.

Fast accurate predictions of 1H NMR spectra from the molecular structure of organic compounds are much in demand for structure validation, automatic structure elucidation, or for the calibration of chemometric methods with simulated data.

The SPINUS software package is a neural network system mostly developed in my group for the prediction of 1H NMR properties from the molecular structure. It was trained using databases of experimental NMR chemical shifts, and relies on a series of empirical descriptors representing the physicochemical, topological and geometrical environments of hydrogen atoms in a molecule. Ensembles of feed-forward neural networks (FFNN) were incorporated into Associative Neural Networks (ASNN), which correct a prediction obtained by the FFNNs with the observed errors for the k nearest neighbours in an additional memory. We have recently reported a procedure to also estimate coupling constants with the ASNNs trained for chemical shifts. The SPINUS program is now able to simulate the full 1H NMR spectrum from the molecular structure. Its implementation as a web service can be accessed at http://www.dq.fct.unl.pt/spinus.

Refs:

1. Y. Binev, J. Aires-de-Sousa, J. Chem. Inf. Comput. Sci. 2004, 44, 940.
2. Y. Binev, M. Corvo, J. Aires-de-Sousa, J. Chem. Inf. Comp. Sci. 2004, 44, 946.
3. F. B. Da Costa, Y. Binev, J. Gasteiger, J. Aires-de-Sousa, Tetrahedron Lett. 2004, 45, 6931.
4. Y. Binev, M. M. B. Marques, J. Aires-de-Sousa, J. Chem. Inf. Model. 2007, 47(6), 2089-2097.