Starting Aug. 16, 2025 at 9:00 AM (Eastern Standard Time), this server will be shut down temporarily to migrate services to our new server at
https://colmar.cloud. Please update your bookmarks accordingly.
All saved user jobs, files, and settings will be migrated automatically. The new server will be fully operational by
Aug. 18, 2025.
This current server will remain accessible until Aug. 31, 2025. However, any jobs submitted after
Aug. 16, 2025 at 9:00 AM (EST) will NOT be migrated to the new server.
Deep Neural Network and Voigt Fitter for Accurate Deconvolution of Complex 1D NMR Spectra
Step 1 option 1 Time domain data uploading
Bruker fid data format information
You can find raw fid data in the experimental folder, such as your_exp_name/2, where 2 is EXP ID. The file is usually named fid (or ser)
The acquisition status file is usually called acqus (not acqu) in the same folder as the FID data file.
You can also drag and drop fid and/or acqus file, or their parent folder to the box above.
The server will run automatic phase correction, after ZF and FF.
Step 1 option 2, Processed ( frequency domain) data uploading
Please apply zero filling in your processing using nmrPipe command "-zf 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” to achieve desired peak shape (Voigt).
Beside nmrPipe .ft1 format, you can also use Topspin text file generated using command "totxt" or CSV format from some software (two columns by space, ppm and amplitude).
You need to keep the imaginary part if you want to run manual phase correction.
Please adjust P0 first, then P1. Anchor point will be set to center of the visible region of the spectrum. Click on "Apply P1" to actually apply the phase correction.
Step 3 Baseline correction (optional)
A appreciate baseline correction can only be defined by the user, based on the experimental conditions. That is, the user need to define the smoothness parameter
and positive spectrum parameter then apply the baseline correction after satisfying result.
The smoothness parameter is the weight of the smoothness constraint in the baseline estimation. The larger the value, the smoother the baseline. Typical values are from 1e8 to 1e13
The positive spectrum parameter is the weight of the constraint that the baseline should be below the spectrum. The larger the value, the less the baseline will be allowed above the spectrum (to accommodate noise). Typical values are from 1.0 to 2.0
Important: phase correction is no longer allowed after baseline correction.
Reference: Baseline Correction for NMR Spectroscopic Metabolomics Data Analysis (Yuanxin Xi and David M Rocke); BMC Bioinformatics 2008, 9:324
Step 4 Spectrum deconvolution
DEEP Picker model selection
Model 1 was optimized for peak width (defined as Full Width as Half Height) from 8 to 20, with 12 points as optimal value
Model 2 was optimized for peak width from 4 to 12, with 8 points as optimal value
The default model, DEEP Auto, will estimate median peak width, then run cubic spline interpolation to adjust it to 6.0 and apply model 2.
For spectrum with very heterogeneous peak width, set a compromise interpolation step (for model 2). A step of 0.5 means spectral size will be doubled.
(optional) Pseudo 2D fitting of multiple spectra
This plot supports pan and zoom operations. You can also pan or zoom on single X or Y axis by setting cursor on X or Y axis.
Deconvoluted peaks will only show when zoomed in to certain scale. Black: experiment; Red: Reconstruction; Green: deconvoluted peaks.
Magenta: total spectrum from all matched compounds
This plot supports pan and zoom operations. You can also pan or zoom on single X or Y axis by setting cursor on X or Y axis.
Deconvoluted peaks will only show when zoomed in to certain scale. Black: experiment; Red: Reconstruction; Green: deconvoluted peaks.
Magenta: total spectrum from all matched compounds
Reference
