Advanced Deconvolution Parameters
The basic Basic Deconvolution Parameters page and the Quick Controls page discussed some of the key parameters UniDec uses for deconvolution. These are great features and are good enough in the majority of cases. However, there are a range of other useful tricks that we can use for improving deconvolution, especially in tricky cases. These are found in the Advanced Deconvolution Parameters yellow tab on the main control panel.
As described in the Quick Controls page page, the peak width defined for the deconvolution can be a very important aspect of the deconvolution. There are tools for detecting the peak width (see the Quick Controls page). Here, you can manually edit the peak FWHM and the peak shape. The Gaussian and Lorentzian peak shapes are standard. The "Split G/L" has a Gaussian peak shape on the low mass side and a Lorentzian peak shape on the high mass side. They are matched to be symmetric at the half maximum, but the Lorentzian tail will then extend further at higher mass. This asymmetric peak simulates adduction common in native MS.
Next, you can precisely control the Beta, Charge Smooth Width, Point Smooth Width, and Mass Smooth Width that are operated by the quick controls. There are a few hidden features here that may be worth exploring. Using negative values for the Mass and Charge Smooth Widths will use a different smoothing function for the mass and charge smoothing/filtering (see How does UniDec Work?). The positive values use a log mean filter. Negative values use a Gaussian filter. The negative values can be very useful when there is a broad charge or mass distribution. Feel free to explore!
Next, you can limit the number of iterations on the algorithm. UniDec will naturally stop if it reaches a sufficient convergence, but you can force it to stop sooner by lowering this. It will likely speed things up, but it may sacrifice deconvolution quality.
The m/z to Mass transformation controls how the UniDec converts the m/z axis (which can be linear or nonlinear, see Data Processing) into the linear mass axis. It can "Integrate" all the m/z bins into the nearest mass bin, which works well if the mass data is undersampled compared to the m/z data. It can also "Interpolate" the m/z axis and find the interpolated value for each mass as it is projected back on the m/z axis, which works well if the mass data is oversampled compared to the m/z data. In other words, the integration is forward looking to force m/z into mass space. The interpolation is backward looking to force the mass space back into the m/z space. The "Smart" mode automatically senses whether the mass is over or under sampled compared to m/z and selects the appropriate option. It actually operates on a point-by-point basis, so it can interpolate part of the spectrum and integrate another part depending on the sampling rate in m/z. The only reason to not use "Smart" mode would be if you wanted to be highly quantitative and force integration. In integration mode, the area under the spectrum will be perfectly preserved.
The Adduct Mass setting described the mass of the electrospray adduct. In positive mode, this is almost always a proton. In negative mode, it should be the negative mass of a proton. There is a button below this to select negative mode, which just automatically sets this value. If you are using crazy sodium adduct samples or something, you could adjust this, but most users will only use plus or minus protons. Note, in Version 5.2.1 and later, UniDec will attempt to automatically read the polarity from Raw and mzML data types. However, you should still check this.
Finally, charge scaling is used to correct some biases in Orbitrap and FT-ICR detection. For each of these techniques, the signal is proportional to the charge state, so higher charged ions will give a larger signal than lower charged ions at the same concentration. Checking the "Charge Scaling" box will divide each signal intensity by its charge state to give signal intensities that scale better with concentration rather than charge. This is mostly useful for super quantitative applications.
The next few features help refine your mass and charge space in a more sophisticated way than with the basic parameters. The first way is with "Manual Mode". Manual mode allows you to select specific regions of the spectrum and assign the charge state manually to a single charge. You can open the tool to set these with "Tools > Manual Assignment". This will let you zoom into specific regions of the m/z spectrum and define the charge. Once you have set the m/z ranges and charge states, you need to turn on Manual Mode by clicking the button. It will then use these manual assignments in the deconvolution. Manual mode is very useful in cleaning up complex spectra. You can perform an initial unbiased deconvolution, manually comb through the results, and assign the peaks that you are confident in. UniDec will these figure out the other peaks from there. Iterating through this process can help figure out complex spectra with some expert user guidance.
Another way to limit the masses is to set a "Mass List". The mass list can be set through the left panel of the "Tools > Oligomer and Mass Tools". Here, you can either manually enter or automatically calculate a list of masses. Selecting the "Mass List Window" option will then only allow masses that are with in the "Window" of that peak. For example, if you had a series of peaks around 50 kDa and another series around 150 kDa, you could enter each of these masses in the list, set an appropriate window to capture the masses you want (say 10 or 20 kDa), and filter out anything in between that is artifactual.
Finally, you can also filter by the native charge. Prior studies of globular proteins have shown that there is a consistent relationship between mass and the average charge state under native MS conditions that can be described as: nativez = 0.0467 * (mass ** 0.533). There are a variety of reasons your charge states may be higher (unfolding) or lower (charge reduction or CID), but this relationship can be a useful one to filter out charge state that are too high or too low for their respective masses. The upper and lower limits for the deviation from native charge can be set in the "Native Charge Offset Range". This feature can be useful for cutting of high charge states of small things while still allowing larger thing to have higher charge states. If you are interested in playing more with the native charge and visualizing your native charge offset, try "Analysis > Native Charge/Mass Tools" and "Analysis > Plot Charge Offsets".
Isotope mode is a tricky beast, and I definitely consider it experimental. Turning it on with "Average" isotope mode is safest. This will use the isotope distributions to help deconvolve the data. It will plot the data in normal average isotopic mode, as you would see in the spectrum. Using "Monoisotopic" mode will try to not only use the isotopes for deconvolution but also attempt to deisotope them, meaning showing you a single peak at the monoisotopic mass for each cluster of isotopes. It uses an averagine model, so it only works for proteins. Also, it's common that to find +/- 1 artifacts in monoisotopic mode or more. In general, my recommendation is to use the "Isotopic Resolution" preset as a place to start and to use average mode where possible. But, feel free to explore and try it yourself. The implementation of isotope mode is actually really cool if you want to peak behind the scenes (if you are reading this far, you are clearly a super user and amazing human being), but it is just a challenging problem with a lot of potential artifacts. That said, average mode is relatively robust and worth a try if your data is isotopically resolved.