Key points are not available for this paper at this time.
To extract information from NMR experiments, users need to identify the number of resonances in the spectrum, together with characteristic features such as chemical shifts and intensities. In many applications, particularly those involving biomolecules, this procedure is typically a manual and laborious process. While many algorithms are available to tackle this problem, their performance tends to be inferior to that of an experienced user. Here, we introduce UnidecNMR, which identifies resonances in NMR spectra using deconvolution. We demonstrate its favourable performance on 1 and 2D simulated spectra, strongly overlapped 1D spectra of oligosaccharides and 2D HSQC, 3D HNCO, 3D HNCA and 3/4D methyl-methyl NOE experimental spectra from a range of proteins. UnidecNMR outperforms a number of freely available algorithms and provides results comparable to those generated manually. Introducing additional restraints, such as a 2D peak list when analysing 3 and 4D data and incorporating reflection symmetry in NOE analysis further improves the results. UnidecNMR outputs a back-calculated spectrum and a peak list, both of which can be easily examined using the supplied GUI. The software allows interactive processing using nmrPipe, allowing users to go directly from raw data to processed spectra with picked peak lists.
Buchanan et al. (Tue,) studied this question.