Peak assignment in automatic data analysis.

Linear prediction algorithms are able to identify peaks in an NMR spectrum, but are not able to assign these peaks to components anticipated in the spectrum. We have developed an artificial-intelligence protocol which uses the output parameter list from an LPSVD algorithm, and automatically assigns the peaks on the basis of an anticipated list of components. To overcome the influence of experimental conditions on the absolute values of frequency, integrated area, and linewidth, the assignment routine performs an internal scaling of the data by comparing all possible pairs of peaks in the spectrum. Completely automated analysis of large numbers of in vivo FIDs is now possible.