Deep Neural Network for Accurate Deconvolution of Complex Multidimensional NMR Spectra


Import spectrum processing requirements, pleas read carefully.

1. The number of points per peak (PPP), which is given by the number of data points that sample the peak’s full width at half height (FWHH), is allowed to vary either from 6 to 20 points or from 4 to 12 points, whereby the former is typical for protein and the latter for metabolomics spectra. The Lorentzian component can vary from 0%, corresponding to a pure Gaussian lineshape, to 100%, corresponding to a pure Lorentzian lineshape. Typically, 3 fold of zero-filling is required to reach these PPPs.

2. Please use the 2pi-Kaiser window function for apodization in the NMRPipe software with the command “SP -off 0.5 -end 0.896 -pow 3.684”. The resulting peak shapes can then be modeled in all dimensions with good accuracy using a Voigt profile that combines variable amounts of Lorentzian vs. Gaussian character.

3. When you adjust lowest contour level, peaks that are fall below it will become invisible. However, all peaks will be kept in the downloaded file.

STEP1, Spectrum Uploading and Visualization:

Your name and institute*
HSQC spectrum file
(Notice: spectra files can be nmrpipe (.ft2), sparky (.ucsf), Topspin ASCII format (.txt)) or Mnova .csv format. The server use file extension to distinguish file format!

Load previous session:
You previous session id

(You can type in "asynu" or "urine" into session id box to load in pre-picked examples.)

STEP2
Minimal amplitude for picked peaks: times of estimated noise level
ANN model selection, FWHH: 6-20 or 4-12
Download picked peak list file in pipe format

Peak shape: Gaussian Voigt
Minimal amplitude for picked peaks: times of estimated noise level
Max number of rounds in non-linear fitting:
Download reconstructed spectrum file in pipe format
Download fitted peak list file in pipe format


Estimated noise level of spectrum is

*This information is only used for the compilation of our webserver usage data.

This server is last updated at February 23 2021 09:49:04.