-
Notifications
You must be signed in to change notification settings - Fork 22
Quick Controls
This section will cover the quick control parameters in UniDec and the deconvolution parameters they control.
The charge smooth function is a filter used in the first step of the UniDec algorithm. Briefly, this tells UniDec that you expect a smooth charge state distribution for each species. Therefore, UniDec will expect that a particular peak with a charge state of z should also have adjacent peaks with charge states at z+1 and z-1. The width that you enter tells UniDec how many charge states to consider on each side, but a value of 1 is usually optimal. That is why the Quick Controls presents it as a simple check box for whether you expect a smooth charge state distribution or not. However, you can adjust the charge smooth width parameter to fine tune the deconvolution, which is helpful in some cases. Note: setting this parameter does not force the algorithm into a Gaussian charge state distribution or any particular model but only biases it towards smoother charge state distributions.
In more detail, charge smoothing uses a mean filter (also called a rolling average or boxcar filter) in log space throughout the charge state distribution for a given mass. Checking the “Smooth Charge State Distributions” box will automatically set the width of the charge state smoothing to 1. The width specifies how many data points to include in the smoothing (+/- n). For example, if a width of 1 is used (which is the default), for each potential mass value, the log intensity for charge states z, z-1, and z+1 will be averaged and set as the intensity at data point z for that mass. The log intensities are used to prevent broadening of the charge state distribution that can otherwise occur if the mean filter was used on the normal intensity values.
One hidden feature: If a negative value is set for the charge smooth function, it will change how the charge state distribution is smoothed. Rather than using a logarithmic mean filter, a negative value will tell UniDec to smooth the charge state distribution with a simple Gaussian filter that has a width set by the value in the box. This Gaussian smoothing will broaden the charge state distribution more than the logarithmic mean function. Using a larger filter width will push the charge state distributions to be wider. When choosing which charge state smoothing filter to use, consider whether you would like a narrower charge state distribution (positive) or a wider charge state distribution (negative). Normally, we recommend keeping the charge smooth at 1 but experimenting with this parameter can help deconvolve complex mass spectra.
Clicking the “Automatic m/z peak width” option will tell UniDec to determine the peak shape and FWHM from the processed mass spectrum. UniDec does this by finding the most abundant peak, isolating it, and fitting it to either a Gaussian, Lorentzian, or split Gaussian/Lorentzian peak and then finding the best fit from these three. These parameters can also be manually set in the additional deconvolution parameters tab. Under Tools > Peak width tool, you can select and fit different peaks manually. Clicking OK will update the tab. You can return to the automatic peak shape by selecting Tools > Automatic Peak Width or Ctrl + W. It’s better to use the lowest FWHM possible to capture all of the features in the data, but lower FWHM values can lead to artifacts in the deconvolution. A higher FWHM can lead to underfitting the data, resulting in a less well resolved deconvolved mass spectra that is cleaner but does not capture the fine details present in the spectrum. Overall, the FWHM can be a very useful parameter to tune the deconvolution.
The smooth nearby points option affects the point smooth function in the additional deconvolution parameters tab. Briefly, smoothing nearby points tells UniDec that neighboring data points should have the same charge state, which is usually true in continuously sampled data. In more detail, this function will smooth neighboring data points within each charge state with a mean filter. The width of data points to smooth is set by the value in the box. There are 4 options for the smooth nearby points function in the Quick Controls: none, some, lots, and other. Clicking the “some” option will turn the point smooth function on and set it to the default, which is 1. The “lots” option will set the point smooth to 10 and should be used sparingly, because using a wider point smooth width may broaden the peak widths slightly. Using point smooth other than 1 or 10 will switch the smooth nearby points option to other.
The suppress artifacts feature sets the beta parameter in the additional deconvolution parameters tab. The beta parameter is useful for deconvolving mass spectra that contain broad mass peaks and for reducing artifacts in the data. In these cases, it can be difficult to assign one unique charge state for each mass peak, resulting in artifact peaks in the deconvolved mass spectrum. The beta parameter uses a SoftMax function to push each data point towards having a single charge state. It will push the most abundant charge state in the 2-D m/z and charge matrix for each potential mass to be more intense. The amount of increased intensity is determined by the beta parameter. The 4 options for suppress artifacts: none, some, lots, and other. Clicking the “some” option will set the beta value at 50, which is a medium setting that will suppress some artifacts. Clicking “lots” will set the beta value at 500, which will remove more artifacts but also may start to distort the deconvolution, especially when there are overlapping peaks in the spectrum. Using lower values in the range of 5 to 15 can also be useful for mild cleanup of deconvolution. To learn more about the beta function refer to: Marty, M. T., Eliminating Artifacts in Electrospray Deconvolution with a SoftMax Function. J. Am. Soc. Mass. Spectrom. 2019, 30 (10), 2174-2177.
Similar to charge smoothing, mass difference is an additional filter that can be applied in the first step in the UniDec algorithm. Mass difference suggests that each m/z should have a neighboring peak that is (m+nM)/z and (m-nM)/z, where M is the value in Da entered in the box. For example, M could be a repeating monomer unit of a polymer or a repeating lipid mass in a nanodisc. Mass difference will help UniDec identify the repeating masses by using the input mass spacing as a prior assumption to help assign the charge states. Furthermore, in the Additional Deconvolution Parameters tab, mass smooth width can be used to set how many neighboring mass peaks to consider. In other words, it sets how large n should be. By default, the mass smooth width is set to 0, but clicking the box sets it to 1. However, it may be useful to experiment with higher ¬_n_ values when broad mass distributions are expected, which may help smooth the mass distribution and reduce artifacts. Similar to charge state smoothing, setting the value to a negative number activates a hidden feature to switch the filter type from a log mean filter to a Gaussian filter, which can be very beneficial for wider mass distributions.