Binning |
Spectral binning is a widely-used technique where the spectrum is subdivided into several regions (bins), and the total area within each bin is used as an abstracted representation of the original spectrum. The area encapsulated by a bin would ideally capture all of the area associated with a given resonance across all spectra in the dataset, thereby mitigating the effect of minor peak shift and line width variations for a compound across samples.
This command is available by following the menu 'Processing/More Processing/Binning'. Then, the 'Binning' dialog box will open and the user can choose the desired limits of the spectrum to apply the binning (or select Full Spectrum). The user can also select the width of each integral region. A typical 64k point NMR spectrum would be reduced using bin widths of 0.04 ppm, resulting in ~250 bin integral values.
You can also bin 2D datasets:
The 'Sum' method will sum all the points of the bucket, the 'Average Sum' will divide the sum by the number of points on the bin, the 'Center' method will only return the value found in the middle of the bin. For example if the bin has 5 points it will return the value for 3rd point. The 'Peaks' method will use GSD for the peak picking.
There will be an option to save the matrix as CSV. The file generated is the same as the one obtained when saving as "Script: NMR CSV Matrix (Transposed).
Binning (also called bucketing) is a tool used in the multivariate analysis of NMR-based metabonomics data to address the NMR peak misalignment issue. In NMR metabonomics, small variations in the resonance position of the individual peaks caused by experimental and instrumental variations can adversely impact PCA results.
Fixed-width binning (usually at 0.04 ppm) is commonly used to alleviate the impact of the peak misalignment by averaging up the data points falling inside the bin width. However, since it drastically reduces the data resolution, it makes it more difficult to interpret the PCA results such as identifying the changed metabolites from the loadings plot.
The goal of binning a spectrum is to produce a new spectrum in which each new spectral data point corresponds to the integral of a bucket with a given width (e.g. 0.04 ppm). In other words, this is simply a data reduction operation analogous to bucketing integration. Once the spectrum has been binned, you can export it in ASCII format in order to use it for further PCA analysis.
It is often convenient to use the Normalize processing command to normalize the spectrum after applying a binning.
See also here an practical example of this feature |